LUNG CANCER DETECTION ON CT IMAGES BY USING WAVELET … · Web view2019-07-21 · Clustering...

LUNG CANCER DETECTION ON CT IMAGES BY USING WAVELET TRANSFORM

CHAPTER-1

INTRODUCTION

Currently, cancer is still a serious problem in the world. World Cancer 2014

International Agency for Research on Cancer (IARC) at the World Health Organization that

in 2012 there were approximately 14 million new cases of cancer occur. Lung cancer is the

most common cause of death with an estimated 8.2 million deaths. According to IARC, about

70 percent of cancer deaths occur in Africa, Asia, Central and South America. Lung cancer is

a disease caused by lung cell division that is uncontrolled. If left untreated, the growth of lung

cells to grow and spread beyond the lungs. Lung nodules are a circle or oval that is often

found on chest radiograph (X-ray) or computed tomography (CT) scan. At one of the 500

reading chest X-ray can be found a nodule. Benign nodules or benign when found in people

who are still aged under 40 years. In the X-ray will reveal a picture of a tumour or a liquid,

while the CT scans, you will see a tumour, fluid or lymph node enlargement.

Furthermore, to ensure nodules or not to be held laboratory tests of samples of

sputum and phlegm. It is also taken of lung tissue or gland, for laboratory test. Diagnostic

process as above is not effective. Therefore we need software that can automatically detect

lung nodules. The end result is software that can detect an existing image in CT lung nodule

or artery.

1.1 DIGITAL IMAGE PROCESSINGThe identification of objects in an image would probably start with image processing

techniques such as noise removal, followed by (low-level) feature extraction to locate lines,

regions and possibly areas with certain textures.

The clever bit is to interpret collections of these shapes as single objects, e.g. cars on a

road, boxes on a conveyor belt or cancerous cells on a microscope slide. One reason this is an

AI problem is that an object can appear very different when viewed from different angles or

under different lighting. Another problem is deciding what features belong to what object and

which are background or shadows etc. The human visual system performs these tasks mostly

unconsciously but a computer requires skilful programming and lots of processing power to

approach human performance. Manipulating data in the form of an image through several

possible techniques. An image is usually interpreted as a two-dimensional array of brightness

values, and is most familiarly represented by such patterns as those of a photographic print,

Department of Electronics and Communication Engineering Page1


slide, television screen, or movie screen. An image can be processed optically or digitally

with a computer.

To digitally process an image, it is first necessary to reduce the image to a series of

numbers that can be manipulated by the computer. Each number representing the brightness

value of the image at a particular location is called a picture element, or pixel. A typical

digitized image may have 512 × 512 or roughly 250,000 pixels, although much larger images

are becoming common. Once the image has been digitized, there are three basic operations

that can be performed on it in the computer. For a point operation, a pixel value in the output

image depends on a single pixel value in the input image. For local operations, several

neighbouring pixels in the input image determine the value of an output image pixel. In a

global operation, all of the input image pixels contribute to an output image pixel value.

These operations, taken singly or in combination, are the means by which the image is

enhanced, restored, or compressed. An image is enhanced when it is modified so that the

information it contains is more clearly evident, but enhancement can also include making the

image more visually appealing.

An example is noise smoothing. To smooth a noisy image, median filtering can be

applied with a 3 × 3 pixel window. This means that the value of every pixel in the noisy

image is recorded, along with the values of its nearest eight neighbours. These nine numbers

are then ordered according to size, and the median is selected as the value for the pixel in the

new image. As the 3 × 3 window is moved one pixel at a time across the noisy image, the

filtered image is formed.

Another example of enhancement is contrast manipulation, where each pixel's value

in the new image depends solely on that pixel's value in the old image; in other words, this is

a point operation. Contrast manipulation is commonly performed by adjusting the brightness

and contrast controls on a television set, or by controlling the exposure and development time

in printmaking. Another point operation is that of pseudo colouring a black-and-white image,

by assigning arbitrary colours to the gray levels. This technique is popular

in thermograph (the imaging of heat), where hotter objects (with high pixel values) are

assigned one color (for example, red), and cool objects (with low pixel values) are assigned

another color (for example, blue), with other colours assigned to intermediate values.



Recognizing object classes in real-world images is a long standing goal in Computer

vision. Conceptually, this is challenging due to large appearance variations of object

instances belonging to the same class. Additionally, distortions from background clutter,

scale, and viewpoint variations can render appearances of even the same object instance to be

vastly different. Further challenges arise from interclass similarity in which instances from

different classes can appear very similar. Consequently, models for object classes must be

flexible enough to accommodate class variability, yet discriminative enough to sieve out true

object instances in cluttered images. These seemingly paradoxical requirements of an object

class model make recognition difficult. This paper addresses two goals of recognition are

image classification and object detection. The task of image classification is to determine if

an object class is present in an image, while object detection localizes all instances of that

class from an image. Toward these goals, the main contribution in this paper is an approach

for object class recognition that employs edge information only. The novelty of our approach

is that we represent contours by very simple and generic shape primitives of line segments

and ellipses, coupled with a flexible method to learn discriminative primitive combinations.

These primitives are complementary in nature, where line segment models straight contour

and ellipse models curved contour. We choose an ellipse as it is one of the simplest circular

shapes, yet is sufficiently flexible to model curved shapes. These shape primitives possess

several attractive properties. First, unlike edge-based descriptors they support abstract and

perceptually meaningful reasoning like parallelism and adjacency. Also, unlike contour

fragment features, storage demands by these primitives are independent of object size and are

efficiently represented with four parameters for a line and five parameters for an ellipse.

Additionally, matching between primitives can be efficiently computed (e.g., with

geometric properties), unlike contour fragments, which require comparisons between

individual edge pixels. Finally, as geometric properties are easily scale normalized, they

simplify matching across scales. In contrast, contour fragments are not scale invariant, and

one is forced either to rescale fragments, which introduces aliasing effects (e.g., when edge

pixels are pulled apart), or to resize an image before extracting fragments, which degrades

image resolution.

The generic nature of line segments and ellipses affords them an innate ability to

represent complex shapes and structures. While individually less distinctive, by combining a

number of these primitives, we empower a combination to be sufficiently discriminative.

Here, each combination is a two-layer abstraction of primitives pairs of primitives (termed



shape tokens) at the first layer, and a learned number of shape tokens at the second layer. We

do not constrain a combination to have a fixed number of shape-tokens, but allow it to

automatically and flexibly adapt to an object class. This number influences a combination’s

ability to represent shapes, where simple shapes favour fewer shape-tokens than complex

ones. Consequently, discriminative combinations of varying complexity can be exploited to

represent an object class. We learn this combination by exploiting distinguishing shape,

geometric, and structural constraints of an object class. Shape constraints describe the visual

aspect of shape tokens, while geometric constraints describe its spatial layout

(configurations). Structural constraints enforce possible poses/structures of an object by the

relationships (e.g., XOR relationship) between shape-tokens.

1.2 CLASSIFICATION OF IMAGES There are 3 types of images used in Digital Image Processing. They are

1. Binary Image

2. Gray Scale Image

3. Color Image

1.2.1 BINARY IMAGE

A binary image is a digital image that has only two possible values for

each pixel. Typically the two colours used for a binary image are black and white though any

two colours can be used. The colour used for the object(s) in the image is the foreground

colour while the rest of the image is the background colour.

Fig 1.1 Binary image



Binary images are also called bi-level or two-level. This means that each pixel is

stored as a single bit (0 or 1).This name black and white, monochrome or monochromatic are

often used for this concept, but may also designate any images that have only one sample per

pixel, such as grayscale images.

Binary images often arise in digital image processing as masks or as the result of

certain operations such as segmentation, thresholding, and dithering. Some input/output

devices, such as laser printers, fax machines, and bi-level computer displays, can only handle

bi-level images

1.2.2 GRAY SCALE IMAGE

A grayscale Image is digital image is an image in which the value of each pixel is a

single sample, that is, it carries only intensity information. Images of this sort, also known

as black-and-white, are composed exclusively of shades of gray (0-255), varying from black

(0) at the weakest intensity to white (255) at the strongest.

Grayscale images are distinct from one-bit black-and-white images, which in the

context of computer imaging are images with only the two colors, black, and white (also

called bi-level or binary images). Grayscale images have many shades of gray in between.

Grayscale images are also called monochromatic, denoting the absence of

any chromatic variation.

Fig 1.2 Gray scale image

Grayscale images are often the result of measuring the intensity of light at each pixel

in a single band of the electromagnetic spectrum (e.g. infrared, visible light, ultraviolet, etc.),

and in such cases they are monochromatic proper when only a given frequency is captured.

But also they can be synthesized from a full color image; see the section about converting to

greyscale.



1.2.3 COLOUR IMAGE

A (digital) colour image is a digital image that includes colour information for

each pixel. Each pixel has a particular value which determines it’s appearing colour. This

value is qualified by three numbers giving the decomposition of the colour in the three

primary colours Red, Green and Blue. Any colour visible to human eye can be represented

this way. The decomposition of a colour in the three primary colours is quantified by a

number between 0 and 255. For example, white will be coded as R = 255, G = 255, B =

255; black will be known as (R,G,B) = (0,0,0); and say, bright pink will be : (255,0,255).

In other words, an image is an enormous two-dimensional array of colour values, pixels, each

of them coded on 3 bytes, representing the three primary colours. This allows the image to

contain a total of 256x256x256 = 16.8 million different colours. This technique is also known

as RGB encoding, and is specifically adapted to human vision.

Fig 1.3 Colour image

It is observable that our behaviour and social interaction are greatly influenced by emotions

of people whom we intend to interact with. Hence a successful emotion recognition system

could have great impact in improving human computer interaction systems in such a way as

to make them be more user-friendly and acting more human-like.

Moreover, there are a number of applications where emotion recognition can play an

important role including biometric authentication, high-technology surveillance and security

systems, image retrieval, and passive demographical data collections.

It is unarguable that face is one the most important feature that characterises human beings.

By only looking ones’ faces, we are not only able to tell who they are but also perceive a lot



of information such as their emotions, ages and genders.

This is why emotion recognition by face has received much interest in computer vision

research community over past two decades.

CHAPTER-2



LITERATURE SURVEY

All researchers have aim to develop such a system which predict and detect the cancer in its

early stages. Also tried to improve the accuracy of the Early Prediction and Detection system

by pre-processing, segmentation feature extraction and classification techniques of extracted

database. The major contributions of the research are summarized below.

T. Sowmiya, M. Gopi, M. New Begin, L.Thomas Robinson - They described Cancer as the

most dangerous diseases in the world. Lung cancer is one of the most dangerous cancer types

in the world. These diseases can spread worldwide by uncontrolled cell growth in the tissues

of the lung. Early detection of the cancer can save the life and survivability of the patients

who affected by this diseases. In this paper we survey several aspects of data mining

procedures which are used for lung cancer prediction for the patients. Data mining concepts

is useful in lung cancer classification. We also reviewed the aspects of ant colony

optimization (ACO) technique in data mining. Ant colony optimization helps in increasing or

decreasing the disease prediction value of the diseases. This case study assorted data mining

and ant colony optimization techniques for appropriate rule generation and classifications on

diseases, which pilot to exact Lung cancer classifications. In additionally to, it provides basic

framework for further improvement in medical diagnosis on lung cancer.

Ada¹, Rajneet Kaur² (2013) – Here it uses a computational procedure that sort the images

into groups according to their similarities. In this paper Histogram Equalization is used for

pre-processing of the images and feature extraction process and neural network classifier to

check the state of a patient in its early stage whether it is normal or abnormal. After that we

predict the survival rate of a patient by extracted features. Experimental analysis is made with

dataset to evaluate the performance of the different classifiers. The performance is based on

the correct and incorrect classification of the classifier. In this paper Neural Network

Algorithm is implemented using open source and its performance is compared to other

classification algorithms. It shows the best results with highest TP Rate and lowest FP Rate

and in case of correctly classification, it gives the 96.04% result as compare to other

classifiers.

Dasu Vaman Ravi Prasad (2013) – Here image quality and accuracy is the core factors of

this research, image quality assessment as well as improvement are depending on the

enhancement stage where low pre-processing techniques is used based on Gabor filter within

Gaussian rules. Following the segmentation principles, an enhanced region of the object of



interest that is used as a basic foundation of feature extraction is obtained. Relying on general

features, a normality comparison is made. In this research, the main detected features for

accurate images comparison are pixels percentage and mask-labelling.

S Vishukumar K. Patela and Pavan Shrivastavab (2012) - In this paper authors mostly

focus on significant improvement in contrast of masses along with the suppression of

background tissues is obtained by tuning the parameters of the proposed transformation

function in the specified range. The manual analysis of the sputum samples is time

consuming, inaccurate and requires intensive trained person to avoid diagnostic errors. The

segmentation results will be used as a base for a Computer Aided Diagnosis (CAD) system

for early detection of cancer, which improves the chances of survival for the patient. In this

paper, authors proposed gabor filter for enhancement of medical images. It is a very good

enhancement tool for medical images.

Fatma Taher1,*, Naoufel Werghi1, Hussain Al-Ahmad1, Rachid Sammouda2 (2012)

This paper presents two segmentation methods, Hopfield Neural Network (HNN) and a

Fuzzy C-Mean (FCM) clustering algorithm, for segmenting sputum color images to detect the

lung cancer in its early stages. The manual analysis of the sputum samples is time consuming,

inaccurate and requires intensive trained person to avoid diagnostic errors. The segmentation

results will be used as a base for a Computer Aided Diagnosis (CAD) system for early

detection of lung cancer which will improve the chances of survival for the patient. However,

the extreme variation in the gray level and the relative contrast among the images make the

segmentation results less accurate, thus we applied a thresholding technique as a pre-

processing step in all images to extract the nuclei and cytoplasm regions, because most of the

quantitative procedures are based on the nuclear feature. The thresholding algorithm

succeeded in extracting the nuclei and cytoplasm regions. Moreover, it succeeded in

determining the best range of thresholding values. The HNN and FCM methods are designed

to classify the image of N pixels among M classes. In this study, we used 1000 sputum color

images to test both methods, and HNN has shown a better classification result than FCM, the

HNN succeeded in extracting the nuclei and cy-toplasm regions. In this paper authors uses a

rule based thresholding classifier as a pre-processing step. The thresh-olding classifier is

succeeded in solving the problem of in-tensity variation and in detecting the nuclei and

cytoplasm regions, it has the ability to mask all the debris cells and to determine the best rang

of threshold values. Overall, the thresholding classifier has achieved a good accuracy of 98%

with high value of sensitivity and specificity of 83% and 99% respectively



CHAPTER-3

PROPOSED SYSTEM



With the advances in imaging technology, diagnostic imaging has become an indispensable

tool in medicine today. X-ray angiography (XRA), magnetic resonance angiography (MRA),

magnetic resonance imaging (CT), computed tomography (CT), and other imaging modalities

are heavily used in clinical practice. Such images provide complementary information about

the patient. While increased size and volume in medical images required the automation of

the diagnosis process, the latest advances in computer technology and reduced costs have

made it possible to develop such systems.

Lung tumor detection on medical images forms an essential step in solving several practical

applications such as diagnosis of the tumors and registration of patient images obtained at

different times. Segmentation algorithms form the essence of medical image applications

such as radiological diagnostic systems, multimodal image registration, creating anatomical

atlases, visualization, and computer-aided surgery

Tumor segmentation algorithms are the key components of automated radiological

diagnostic systems. Segmentation methods vary depending on the imaging modality,

application domain, method being automatic or semi-automatic, and other specific factors.

There is no single segmentation method that can extract vasculature from every medical

image modality. While some methods employ pure intensity-based pattern recognition

techniques such as thresholding followed by connected component analysis, some other

methods apply explicit tumor models to extract the tumor contours. Depending on the image

quality and the general image artifacts such as noise, some segmentation methods may

require image preprocessing prior to the segmentation algorithm. On the other hand, some

methods apply post-processing to overcome the problems arising from over segmentation.

Medical image segmentation algorithms and techniques can be divided into six main

categories, pattern recognition techniques, model-based approaches, tracking-based

approaches, artificial intelligence-based approaches, neural network-based approaches, and

miscellaneous tube-like object detection approaches.

Pattern recognition techniques are further divided into seven categories, multi-scale

approaches, skeleton-based approaches, region growing approaches, ridge-based approaches,

differential geometry-based approaches, matching filters approaches, and mathematical

morphology schemes.

Clustering analysis plays an important role in scientific research and commercial application.

This thesis provides a survey of current tumour segmentation methods using clustering



approach and provides both early and recent literature related to tumour segmentation

algorithms and techniques.

3.1 REGION GROWING APPROACH

Region growing technique segments image pixels that are belong to an object into regions.

Segmentation is performed based on some predefined criteria. Two pixels can be grouped

together if they have the same intensity characteristics or if they are close to each other. It is

assumed that pixels that are closed to each other and have similar intensity values are likely

to belong to the same object. The simplest form of the segmentation can be achieved through

thresholding and component labeling. Another method is to find region boundaries using

edge detection. Segmentation process, then, uses region boundary information to extract the

regions. The main disadvantage of region growing approach is that it often requires a seed

point as the starting point of the segmentation process. This requires user interaction. Due to

the variations in image intensities and noise, region growing can result in holes and over

segmentation. Thus, it sometimes requires post-processing of the segmentation result.

Existing method

Partial derivatives.

Wavelet based denoising.

Thresholding and K means clustering methods for segmentation.

Drawbacks

Loss of edge details.

In wavelet denoising, failure to detect edge details at curved region.

K means - It is not suitable for all lighting condition of images.

Difficult to measure the cluster quality.

3.2 BLOCK DIAGRAM OF PROPOSED SYSTEM



Fig 3.1 Block diagram

Image Pre-processing

Firstly, in image pre-processing median filter is used on grayscale image of CT scan images.

Some noises are embedded on CT Images at the time of image acquisition process which aids

in false detection of nodules. Noise may be detected as cancer nodules sometimes. Therefore,

these noises have to be removed for accurate detection of cancer. Median filter removes salt

and pepper noise from the CT images. After median filter, Gaussian filter is implemented. It

smoothes the image and removes speckle noise from image.

Segmentation

This process locates objects or boundaries which help in acquiring the region of interest in

the image. It partitions the image into regions to identify the meaningful information. In lung

cancer detection it segments the cancer nodule from the CT scan image. In the proposed

model watershed segmentation is implemented. Its main feature is that it can separate and

identify the touching objects in the image. This feature helps in proper segmentation of

cancer nodules if it is touching to other false nodules.

Features extraction



In this stage, features like area, perimeter, centroid, diameter, eccentricity and Mean intensity.

These features later on are used as training features to develop classifier.

Classification

This stage classifies the detected nodule as malignant or benign. Support vector machine

(SVM) is used as classifier. It is supervised machine learning method. SVM defines the

function that classifies data into two classes. The function is defined as D(x)=wT xi+ b where

xi are training inputs, wT is m dimensional vector, and b is bias term. Here, i=1…. M.

D(x)=wT xi+ b ≥ 1 for yi=1 D(x)=wT xi+ b ≤-1 for yi=-1

Description

The Project proposes to spot the tumour from CT scanned medical images using multi

clustering model and morphological process.

The segmentation refers to the process of partitioning a digital image into multiple segments.

The brain CT is taken, and its noises are removed using filters and then applied spatial Fuzzy

C means Clustering algorithm for the segmentation of CT brain images.

The morphological process will be used to smooth the tumour region from the noisy

background.

The segmented primary and secondary regions are compressed with hybrid techniques for

telemedicine application.

3.3 PREPROCESSINGImage restoration is the operation of taking a corrupted/noisy image and estimating

the clean original image. Corruption may come in many forms such as motion blur, noise,

and camera misfocus. Image restoration is different from image enhancement in that the

latter is designed to emphasize features of the image that make the image more pleasing to

the observer, but not necessarily to produce realistic data from a scientific point of view.

Image enhancement techniques (like contrast stretching or de-blurring by a nearest neighbour

procedure) provided by "Imaging packages" use no a priori model of the process that created

the image. With image enhancement noise can be effectively be removed by sacrificing

some resolution, but this is not acceptable in many applications. In a Fluorescence

Microscope resolution in the z-direction is bad as it is. More advanced image processing

techniques must be applied to recover the object. Deconvolution is an example of image



restoration method. It is capable of: Increasing resolution, especially in the axial direction

removing noise increasing contrast.

3.3.1 Dual-tree complex wavelet transforms (DT-CWT)

The dual-tree complex wavelet transform (CWT) is a relatively recent enhancement to

the discrete wavelet transform (DWT), with important additional properties: It is nearly shift

invariant and directionally selective in two and higher dimensions. It achieves this with a

redundancy factor of only 2d for d-dimensional signals, which is substantially lower than the

undecimated DWT. The multidimensional (M-D) dual-tree CWT is no separable but is based

on a computationally efficient, separable filter bank (FB). The theory behind the dual-tree

transforms shows how complex wavelets with good properties can be designed, and

illustrates a range of applications in signal and image processing.

Fig 3.2 The value of the wavelet coefficient in “Real-Valued Discrete Wavelet

Transform and Filter Banks

In the neighbourhood of an edge, the real DWT produces both large and small wavelet

coefficients. In contrast, the (approximately) analytic CWT produces coefficients whose

magnitudes are more directly related to their proximity to the edge. Here, the test signal is a

step edge at n = no, x(n) = u(n − no). The figure shows the value of the wavelet coefficient

d(0, 8) (the eighth coefficient at stage 3 in “Real-Valued Discrete Wavelet Transform and

Filter Banks, as a function of no. In the top panel, the real coefficient d(0, 8) is computed

using the conventional real DWT. In the lower panel, the complex coefficient (0, 8) is

computed using the dual-tree CWT.



3.3.2 Wavelet transform and multistate analysis

The wavelet transform has been exploited with great success across the gamut of signal

processing applications, in the process, often redefining the state of the art performance. In a

nutshell, the DWT replaces the infinitely oscillating sinusoidal basis functions of the Fourier

transform with a set of locally oscillating basis functions called wavelets. In the classical

setting, the wavelets are stretched and shifted versions of a fundamental, real-valued band

pass wavelet ψ(t ). When carefully chosen and combined with shifts of a real-valued low-pass

scaling function φ(t ), they form an orthonormal basis expansion for one-dimensional (1-D)

real-valued continuous-time signals. That is, any finite energy analog signal x(t ) can be

decomposed in terms of wavelets and scaling functions via

The scaling coefficients c(n) and wavelet coefficients d( j, n) are computed via the inner

products,

They provide a time-frequency analysis of the signal by measuring its frequency content

(controlled by the scale factor j) at different times (controlled by the time shift n).There exists

a very efficient, linear time complexity algorithm to compute the coefficients c(n) and d( j, n)

from a fine-scale representation of the signal (often simply N samples) and vice versa based

on two octave-band, discrete-time FBs that recursively apply a discrete-time low-pass filter

h0(n), a high-pass filterh1(n), and up-sampling and down-sampling operations. These filters

provide a convenient parameterization for designing wavelets and scaling functions with

desirable properties, such as compact time support and fast frequency decay (to ensure the

analysis is as local as possible in time frequency) and orthogonality to low-order

polynomials (vanishing moments)



This corresponds to a rotation of both filters in the z-plane by 90°. If h0(n) and h1(n)

satisfy the PR conditions, then so will hp(n) and hn(n). The given low-pass/high-pass

filtersh0(n), h1(n) illustrated in the frequency domain, the complex filters hp(n), hn(n) are

illustrated in the frequency domain in Figure 4. When used by itself, this complex can

effectively separate the positive and negative frequency components of a signal; in a discrete-

time sense, hp(n) and hn(n) are approximately analytic.

Fig 3.3 Analysis FB for the DWT with invertible complex post-filtering.

3.3.3 Shift Variance

A small shift of the signal greatly perturbs the wavelet coefficient oscillation

pattern around singularities. Shift variance also complicates wavelet-domain processing

algorithms must be made capable of coping with the wide range of possible wavelet

coefficient patterns caused by shifted singularities

Fig 3.4 A q-shift complex wavelet corresponding to a set of orthonormal dual-tree filters of

length



To better understand wavelet coefficient oscillations and shift variance, consider a piecewise

smooth signal x(t−t0) like the step function

Analysed by a wavelet basis having a enough vanishing moments. Its wavelet coefficients

consist of samples of the step response of the wavelet

Where is the height of the jump: Since ψ(t ) is a band pass function that oscillates around

zero, so does its step response( j, n) as a function of n (recall Figure 1). Moreover, the

factor2 j in the upper limit (j≥ 0) amplifies the sensitivity of d( j, n)to the time shift t0,

leading to strong shift variance.

3.3.4 Complex WaveletsThe key is to note that the Fourier transform does not suffer from these problems.

First, the magnitude of the Fourier transform does not oscillate positive and negative but

rather provides a smooth positive envelope in the Fourier domain. Second, the magnitude of

the Fourier transform is perfectly shifting invariant, with a simple linear phase offset

encoding the shift. Third, the Fourier coefficients are not aliased and do not rely on a

complicated aliasing cancellation property to reconstruct the signal and fourth, the sinusoids

of them-D Fourier basis are highly directional plane waves. The DWT, which is based on

real-valued oscillating wavelets, the Fourier transform is based on complex-valued oscillating

sinusoids.

The oscillating cosine and sine components (the real and imaginary parts, respectively) form

a Hilbert transform pair; i.e., they are 90◦ out of phase with each other. Together they



constitute an analytic signal ejtthat is supported on only one-half of the frequency axis (_

>0).The oscillating cosine and sine components (the real and imaginary parts, respectively)

form a Hilbert transform pair; i.e., they are 90◦ out of phase with each other. Together they

constitute an analytic signal ejtthat is supported on only one-half of the frequency axis (_

>0).

3.4 CLUSTERINGClustering can be considered the most important unsupervised learning problem, so, it deals

with finding a structure in a collection of unlabeled data. A cluster is therefore a collection of

objects which are “similar” between them and are “dissimilar” to the objects belonging to

other clusters

Clustering algorithms may be classified as listed below

Exclusive Clustering

Overlapping Clustering

Hierarchical Clustering

Probabilistic Clustering

In the first case data are grouped in an exclusive way, so that if a certain datum belongs to a

definite cluster then it could not be included in another cluster. On the contrary the second

type, the overlapping clustering, uses fuzzy sets to cluster data, so that each point may belong

to two or more clusters with different degrees of membership. In this case, data will be

associated to an appropriate membership value. A hierarchical clustering algorithm is based

on the union between the two nearest clusters. The beginning condition is realized by setting

every datum as a cluster. After a few iterations it reaches the final clusters wanted.

3.4.1 K-MEANS CLUSTERING

Cluster analysis, an important technology in data mining, is an effective method of

analysing and discovering useful information from numerous data. Cluster algorithm groups

the data into classes or clusters so that objects within a cluster have high similarity in

comparison to one another but are very dissimilar to objects in other clusters. Dissimilarities

are assessed based on the attribute values describing the objects. Often, distance measures are

used. As a branch of statistics and an example of unsupervised learning, clustering provides

us an exact and subtle analysis tool from the mathematic view K-means algorithm belongs to

a popular partition method in cluster analysis. The most widely used clustering error criterion



is squared-error criterion, it can be defined as

Where J, is the sum of square-error for all objects in the database, xk is the point in space

representing a given object, and mj is the mean of cluster Cj. Adopting the squared-error

criterion, K-means works well when the clusters are compact clouds that are rather well

separated from one another and are not suitable for discovering clusters with no convex

shapes or clusters of very different size. For attempting to minimize the square-emor

criterion, it will divide the objects in one cluster into two or more clusters. In addition to that,

when applying this square-error criterion to evaluate the clustering results, the optimal cluster

corresponds to the extremum. Since the objective function has many local minimal values, if

the result of initialization is exactly near the local minimal point, the algorithm will terminate

at a local optimum. So, random selecting initial cluster centre is easy to get in the local

optimum not the entire optimal. For overcoming that square-error criterion is hard to

Distinguish the big difference among the clusters, one technique has been developed which is

based on representative point-based technique. Besides, there are

Various approaches to solving the problem that the performance of algorithm heavily

depends on the initial starting conditions: the simplest one is repetition with different random

selections some algorithms also employ simulation anneals technique to avoid getting into

local optimal. The idea is that multiple sub-samples are drawn from the dataset clustered

independently, and then these solutions are clustered again respectively, the refined initial

centre is then chosen as the solution having minimal distortion over all solutions. Aiming at

the dependency to initial conditions and the limitation of K-means algorithm that applies the

square-error criterion to measure the quality of clustering, this paper presents a new improved

K-means algorithm that is based on effective techniques of multi-sampling and once-

clustering to search the optimal initial values of cluster centres. Our experimental results

demonstrate the new algorithm can obtain better stability and excel the original K-means in

clustering results.

3.4.2 CLUSTERING MODELClustering can be considered the most important unsupervised learning problem, so, it

deals with finding a structure in a collection of unlabeled data. A cluster is therefore a



collection of objects which are “similar” between them and are “dissimilar” to the objects

belonging to other clusters.

3.4.3 FUZZY CLUSTERING MODEL

Fuzzy clustering plays an important role in solving problems in the areas of pattern

recognition and fuzzy model identification. A variety of fuzzy clustering methods have been

proposed and most of them are based upon distance criteria. One widely used algorithm is the

fuzzy c-means (FCM) algorithm. It uses reciprocal distance to compute fuzzy weights. A

more efficient algorithm is the new FCFM. It computes the cluster centre using Gaussian

weights, uses large initial prototypes, and adds processes of eliminating, clustering and

merging. In the following sections we discuss and compare the FCM algorithm and FCFM

algorithm.

Spatial Fuzzy C Means method incorporates spatial information, and the

membership weighting of each cluster is altered after the cluster distribution in the

neighbourhood is considered. The first pass is the same as that in standard FCM to calculate

the membership function in the spectral domain. In the second pass, the membership

information of each pixel is mapped to the spatial domain and the spatial function is

computed from that. The FCM iteration proceeds with the new membership that is

incorporated with the spatial function. The iteration is stopped when the maximum difference

between cluster centres or membership functions at two successive iterations is less than a

least threshold value.

The fuzzy c-means (FCM) algorithm was introduced by J. C. Bezdek. The idea of FCM is

using the weights that minimize the total weighted mean-square error:

J(wqk, z(k)) = (k=1,K) (k=1,K) (wqk)|| x(q)- z(k)||2

(k=1,K) (wqk) = 1

wqk = (1/(Dqk)2)1/(p-1) / (k=1,K) (1/(Dqk)2)1/(p-1) , p > 1

The FCM allows each feature vector to belong to every cluster with a fuzzy truth value

(between 0 and 1), which is computed using Equation (4). The algorithm assigns a feature

vector to a cluster according to the maximum weight of the feature vector over all clusters.

3.4.4 A New Fuzzy c-means Implementation



Algorithm Flow

Initialize the Fuzzy Weights. In order to comparing the FCM with FCFM, our

implementation allows the user to choose initializing the weights using feature vectors or

randomly. The process of initializing the weights using feature vectors assigns the first K init

(user-given) feature vectors to prototypes then computes the weights by Equation (4).

Standardize the Weights over Q. During the FCM iteration, the computed cluster centres

get closer and closer. To avoid the rapid convergence and always grouping into one cluster,

we use

w[q,k] = (w[q,k] – wmin)/( wmax – wmin)

Before standardizing the weights over Q. Where wmax, wmin are maximum or minimum

weights over the weights of all feature vectors for the particular class prototype.

Eliminating Empty Clusters. After the fuzzy clustering loop we add a step (Step 8) to

eliminate the empty clusters. This step is put outside the fuzzy clustering loop and before



calculation of modified XB validity. Without the elimination, the minimum distance of

prototype pair used in Equation (8) may be the distance of empty cluster pair. We call the

method of eliminating small clusters by passing 0 to the process so it will only eliminate the

empty clusters.

After the fuzzy c-means iteration, for the purpose of comparison and to pick the optimal

result, we add Step 9 to calculate the cluster centres and the modified Xie-Beni clustering

validity

The Xie-Beni validity is a product of compactness and separation measures. The

compactness-to-separation ratio is defined by Equation (6).

= {(1/K)(k=1,K)k2}/Dmin

2

k2 = (q=1,Q) wqk || x(q) – c(k) ||2

Dmin is the minimum distance between the cluster centres.

The Modified Xie-Beni validity is defined as

= Dmin2/ {(k=1,K)k

2 } ]

The variance of each cluster is calculated by summing over only the members of each cluster

rather than over all Q for each cluster, which contrasts with the original Xie-Beni validity

measure.

k2 ={q: q is in cluster k} wqk || x(q) – c(k) ||2

The spatial function is included into membership function as given in Equation

3.5 SEGMENTATION

Image segmentation is a crucial process for most image analysis consequent tasks. Especially,

most of the existing techniques for image description and recognition are highly depend on

the segmentation results. Segmentation splits the image into its constituent regions or objects.

Segmentation of medical images in 2D has many beneficial applications for the medical

professional such as: visualization and volume estimation of objects of concern, detection of



oddities, tissue quantification and organization and many more. The main objective of

segmentation is to simplify and change the representation of the image into something that is

more significant and easier to examine. Image segmentation is usually used to trace objects

and borders such as lines, curves, etc. in images. More accurately, image segmentation is the

process of allocating a label to every pixel in an image such that pixels with the same label

share certain pictorial features. The outcome of image segmentation is a set of segments that

collectively cover the entire image, or a set of edges extracted from the image i.e. edge

detection. In a given region all pixels are similar relating to some distinctive or computed

property, such as texture, intensity or color. With respect to the same characteristics adjacent

regions are significantly different. One of two basic properties of intensity values

Segmentation algorithms are based on: discontinuity and similarity. In the first group we

partition the image based on abrupt changes in intensity, such as edges in an image. The next

group is based on segregating the image into regions that are alike according to a predefined

criterion. Histogram thresholding methodology comes under this group.

3.5.1 K-MEANS SEGMENTATION

K-means is one of the simplest unsupervised learning algorithms that solve the well

known clustering problem. The procedure follows a simple and easy way to classify a given

data set through a certain number of clusters (assume k clusters) fixed a prior. The main idea

is to define k centroids, one for each cluster. These centroids should be placed in a cunning

way because of different location causes different result. So, the better choice is to place them

as much as possible far away from each other. The next step is to take each point belonging

to a given data set and associate it to the nearest centroid. When no point is pending, the first

step is completed and an early group age is done. At this point we need to re-calculate k new

centroids as bray centers of the clusters resulting from the previous step. After we have these

k new centroids, a new binding has to be done between the same data set points and the

nearest new centroid. A loop has been generated. As a result of this loop we may notice that

the k centroids change their location step by step until no more changes are done. In other

words centroids do not move any more. Finally, this algorithm aims at minimizing an

objective function, in this case a squared error function.

3.5.2 HIERARCHICAL SEGMENTATION

A hierarchical set of image segmentations is a set of several image segmentations of

the same image at different levels of detail in which the segmentations at coarser levels of



detail can be produced from simple merges of regions at finer levels of detail. A unique

feature of hierarchical segmentation is that the segment or region boundaries are maintained

at the full image spatial resolution for all segmentations. In a hierarchical segmentation, an

object of interest may be represented by multiple image segments in finer levels of detail in

the segmentation hierarchy and may be merged into a surrounding region at coarser levels of

detail in the segmentation hierarchy. If the segmentation hierarchy has sufficient resolution,

the object of interest will be represented as a single region segment at some intermediate

level of segmentation detail.

A goal of the subject analysis of the segmentation hierarchy is to identify the hierarchical

level at which the object of interest is represented by a single region segment. The object may

then be identified through its spectral and spatial characteristics. Additional clues for object

identification may be obtained from the behaviour of the image segmentations at the

hierarchical segmentation level above and below the level at which the object of interest is

represented by a single region.

3.6 THRESHOLDING

The simplest method of image segmentation is called the thresholding method. This

method is based on a clip-level (or a threshold value) to turn a gray-scale image into a binary

image. The key of this method is to select the threshold value (or values when multiple-levels

are selected). Several popular methods are used in industry including the maximum entropy

method, Otsu's method (maximum variance), and k-means clustering. Recently, methods

have been developed for thresholding computed tomography (CT) images. The key idea is

that, unlike Otsu's method, the thresholds are derived from the radiographs instead of the

(reconstructed) image.

Design Steps

(1) Set the initial threshold T= (the maximum value of the image brightness + the minimum

value of the image brightness)/2.

(2) Using T segment the image to get two sets of pixels B (all the pixel values are less than

T) and N (all the pixel values are greater than T);

(3) Calculate the average value of B and N separately, mean ub and un.

(4) Calculate the new threshold: T= (ub+un)/2



(5) Repeat Second steps to fourth steps up to iterative conditions are met and get necessary

region from the brain image.

3.7ALGORITHM FOR GETTING INITIAL CENTROIDS

Now let's review the standard k-means algorithm. Input: the number of classes and the

population U that

Output: k classes that satisfy the least square error. The process of the algorithm is containing

n objects.

(1) Select k objects randomly from the population U as the initial centroids.

(2) Repeat (3) and (4) until no object changes the class t belongs to.

(3) Compute the distances between each object & and all centroids, and if one object has the

shortest distance from one centroid with rigid to the other centroids then it has the same name

as the centroid all these objects that have the same name belong to the same class.

(4) Average all the vectors of objects belonging to the same class and form the new centroids.

The standard k-means algorithm alternates between assigning the data-points to their closest

centroid (the E-step) and moving each centroid to the mean of its assigned data-points (the

M-step)"'. Because the standard k-means algorithm gets easily trapped in a local minimum

and different initial centroids lead to different results, if we find certain initial centroids that

are consistent with the distribution of data, then a better clustering can be obtained. The aim

of k-means algorithm is to partition objects into several classes and to make the distances

between objects in the same class closer than the distances between objects in different

classes. So, if certain centroids in which each centroid represents a group of similar objects

can be obtained, we will find out the centroids consistent with the distribution of data. Let U

be a data-point set. The initial centroids can be gotten by the following steps. Firstly, compute

the distances between each data-point and all the other data-points in U. Secondly find out the

two data-points between which the distance is the shortest and form a data-point set AI which

contains these two data-points, then we delete them from the population U. Thirdly. compute

distances between each data-point in AI and each data-point in U, find out the data-point that

is closest to the data-point set AI (i.e. of all distances, thy distance between this data-point

and certain data-point in A1 is shortest), delete it from U and add it to AI. Repeat the third

step till the number of data-point in A1 reaches certain threshold. Then we go to step two and

form another data-point set till we get k data-point sets. Finally, the initial centroids can be



gotten by averaging all the vectors in each data-point set.

The algorithm of getting the initial centroids

The Euclidean distance is used in this paper. The distance between one vector

X=(xl,x2,. . .,xn) and the other vector Y=(yl,y2, ...,y n) is described as follows.

The distance between a data-point X and a data-point d(X, V) =min (d(X, Y), Y EV)

Suppose there are n data-points in the population U and we want to partition U into k classes.

Set m=l. Then the algorithm is described as follows. Compute distances between each data-

point and all of the other data-points in U; find the two data points between which the

distance is the shortest and form a data-point set Am (l<m<k) which contains these two data-

points; delete these two data-points from U,

(2) Find the data-point in U that is closest to the data-point set Am, add it to Am and delete it

from U.

(3) Repeat step (2) till the number of data-points in

Am reaches

(4) If m<k, then m=m+l; find another pair of data-points between which the distance is the

shortest in U and form another data-point set Am and delete them from U then go to step (2)

5) For each Am (I<m<k) sum the vectors of data-points and divide the sum by the number of

data-points in Am, then each data-point set outputs a vector and we select these vectors as the

initial centroids.

(6) Execute the process of the standard k-means algorithm from step 2.

The value of a is different about different data. If the value of a is too small, all the centroids

may be obtained in the same region that contains many similar data-points; but if the value of

a is too big, the centroids may stay away from the region that contains many similar data-

points. According to the results of our experiment,

Better clustering can often be obtained if the value of a is set to be 0.75.



CHAPTER-4

TRANSFORM TECHNIQUES

4.1 DISCRETE COSINE TRANSFORM (DCT)

In this lab we use the DCT block from the Video Processing Library in Simulink.

Still, it is useful to understand the various steps in computing a DCT on an image as given

below.

1. First divide the image (N by M pixels) into blocks of 8x8 pixels as shown in Figure.



Fig 4.1 Original Image and Image divided into 8x8 blocks.

2. Compute the DCT on the 8x8 blocks using the formula given below. Please note that in

this lab we use the DCT block from Video Processing Library in Simulink and not the

formula given below.

Formula to calculate DCT on an 8x8 block image.

3. Select the high-frequency components via RxR mask. Please note that both the images

are displayed in a log-amplitude scale in Figures 3.1(c) and 3.1(d) below.

(a) (b)

Fig 4.2(a) 8x8 block image in log amplitude scale

Fig 4.2(b) After applying RxR mask on the image where R=4



4. Compute the 8x8 block IDCT (Inverse DCT) on the compressed DCT coefficients

using the formula given below. Please note that in this lab we use the IDCT block from

the Video and Image Processing Block set in Simulink and not the formula given below.

Formula to calculate IDCT on a 8x8 block image

5. Figures 3.1(a), 3.1(b) show lossy compression results using DCT with R=3, R=2 and

R=1. Please note that R=3 means that we are selecting and transmitting only a 3x3 block

from a 8x8 block resulting in a compression of 86%. As the value of R decreases,

compression increases but picture quality decreases too.

3.1(f)

Fig 4.3(a)Original Image. Fig 4.3(b): R=3 (86% compression).

Figure 4.4(a): R=2 (93% compression). Figure 4.4(b): R=1 (98% compression).

4.2 ENCRYPTION

For encrypting the video we perform additional actions after Step 3. Just for recap, in Step 3



we select an RxR matrix from the computed 8x8 block of DCT coefficients. To add

encryption to the block of RxR DCT co-efficient, we multiply it with a scrambling matrix as

shown in Figure below. Please note that this is element-wise matrix multiplication and not the

usual matrix multiplication.

Scrambling process

4.3 DECRYPTION

For decrypting the video we perform additional actions before Step 4. Just for recap, in Step 4

we apply IDCT on the RxR block of DCT co-efficient. Before computing IDCT we multiply

the scrambled RxR block of DCT co-efficient with the de-scrambling matrix as shown in

Figure. Please note that the scrambling matrix and the descrambling matrix are the same in

our case. Also note that we are doing element-wise matrix multiplication and not the usual

matrix multiplication.

Scrambling and De-scrambling process



4.4 DISADVANTAGES OF DCT

1. Only spatial correlation of the pixels inside the single 2-D block is considered and the

correlation from the pixels of the neighbouring blocks is neglected

2. Impossible to completely decorrelate the blocks at their boundaries using DCT

3. Undesirable blocking artifacts affect the reconstructed images or video frames. (High

compression ratios or very low bit rates)

4. Scaling as add-onàadditional effort

5. DCT function is fixedàcannot be adapted to source data

6. Does not perform efficiently for binary images (fax or pictures of fingerprints)

characterized by large periods of constant amplitude (low spatial frequencies), followed by

brief periods of sharp transitions

4.5 DISCRETE WAVELET TRANSFORM

The wavelet transform (WT) has gained widespread acceptance in signal processing and

image compression. Because of their inherent multi-resolution nature, wavelet-coding

schemes are especially suitable for applications where scalability and tolerable degradation

are important Recently the JPEG committee has released its new image coding standard,

JPEG-2000, which has been based upon DWT.Wavelet transform decomposes a signal into a

set of basis functions. These basis functions are called wavelets. Wavelets are obtained from a

single prototype wavelet y(t) called mother wavelet by dilations and shifting:

Where a is the scaling parameter and b is the shifting parameter

4.5.1 Theory of Wavelet

The wavelet transform is computed separately for different segments of the time-

domain signal at different frequencies. Multi-resolution analysis: analyses the signal at

different frequencies giving different resolutions MRA is designed to give good time

resolution and poor frequency resolution at high frequencies and good frequency resolution



and poor time resolution at low frequencies. Good for signal having high frequency

components for short durations and low frequency components for long duration.e.g. Images

and video frames Theory of WT (cont.) Wavelet transform decomposes a signal into a set of

basis functions. These basis functions are called wavelets.Wavelets are obtained from a single

prototype wavelet y(t) called mother wavelet by dilations and shifting

Where a is the scaling parameter and b is the shifting parameter

4.5.2 1D-WT

The 1-D wavelet transform is given by:

The inverse 1-D wavelet transform is given by:

Discrete wavelet transforms (DWT), which transforms a discrete time signal to a

discrete wavelet representation. it converts an input series x0, x1, ..xm, into one high-pass

wavelet coefficient series and one low-pass wavelet coefficient series (of length n/2 each)

given by:



Where sm(Z) and tm(Z) are called wavelet filters, K is the length of the filter, and i=0, ...,

[n/2]-1.

In practice, such transformation will be applied recursively on the low-pass series until the

desired number of iterations is reached.

4.5.3 Lifting schema of DWT

Lifting schema of DWT has been recognized as a faster approach.

The basic principle is to factorize the polyphase matrix of a wavelet filter into a

sequence of alternating upper and lower triangular matrices and a diagonal matrix. This leads

to the wavelet implementation by means of banded-matrix multiplications.

4.5.4 Two lifting schema



Where si(z) (primary lifting steps) and ti(z) (dual lifting steps) are filters and K is a

constant. As this factorization is not unique, several {si(z)}, {ti(z)} and K are admissible.

4.5.5 2-D DWT for Image



2-D DWT for Image:



4.6 ADVANTAGES OF DWT OVER DCT

1. No need to divide the input coding into non-overlapping 2-D blocks, it has higher

compression ratios avoid blocking artifacts.

2. Allows good localization both in time and spatial frequency domain.

3. Transformation of the whole imageà introduces inherent scaling

4. Better identification of which data is relevant to human perceptionà higher compression

ratio

5. Higher flexibility: Wavelet function can be freely chosen

6. No need to divide the input coding into non-overlapping 2-D blocks, it has higher

compression ratios avoid blocking artifacts.

7. Transformation of the whole imageà introduces inherent scaling

8. Better identification of which data is relevant to human perceptionà higher compression

ratio (64:1 vs. 500:1)

9. Performance

10. Peak Signal to Noise ratio used to be a measure of image quality

11. The PSNR between two images having 8 bits per pixel or sample in terms of decibels

(dBs) is given by:



12. PSNR = 10 log10

13. Mean square error (MSE)

14. Generally when PSNR is 40 dB or greater, then the original and the reconstructed images

are virtually indistinguishable by human observers

4.7 DISADVANTAGES OF DWT

1. The cost of computing DWT as compared to DCT may be higher.

2. The use of larger DWT basis functions or wavelet filters produces blurring and ringing

noise near edge regions in images or video frames

3. Longer compression time

4. Lower quality than JPEG at low compression rates



CHAPTER-5

FEATURE EXTRACTION AND NEURAL NETWORK

5.1 CO-OCCURRENCE MATRIXOriginally proposed by R.M. Haralick, the co-occurrence matrix representation of

texture features explores the grey level spatial dependence of texture]. A mathematical

definition of the co-occurrence matrix is as follows

- Given a position operator P(i,j),

- let Abe an n x n matrix

- Whose element A[i][j] is the number of times that points with grey level

(intensity) g[i] occur, in the position specified by P, relative to points with grey

level g[j].

- Let C be the n x n matrix that is produced by dividing A with the total number of

point pairs that satisfy P. C[i][j] is a measure of the joint probability that a pair of

points satisfying P will have values g[i], g[j].

- C is called a co-occurrence matrix defined by P.

Examples for the operator P are: “i above j”, or “i one position to the right and two below j”,

etc.

This can also be illustrated as follows… Let t be a translation, then a co-occurrence matrix

Ctof a region is defined for every grey-level (a, b) by [1]:

Here, Ct(a, b) is the number of site-couples, denoted by (s, s + t) that are separated by a

translation vector t, with a being the grey-level of s, and b being the grey-level of s + t.

For example; with an 8 grey-level image representation and a vector t that considers only one

neighbour, we would find:

Fig 5.1 Image example



Fig 5.2 Classical Co-occurrence matrix

At first the co-occurrence matrix is constructed, based on the orientation and distance

between image pixels. Then meaningful statistics are extracted from the matrix as the texture

representation .Haralick proposed the following texture features:

1. Energy

2. Contrast

3. Correlation

4. Homogenity

Hence, for each Haralick texture feature, we obtain a co-occurrence matrix. These co-

occurrence matrices represent the spatial distribution and the dependence of the grey levels

within a local area. Each (i,j) th entry in the matrices, represents the probability of going from

one pixel with a grey level of 'i' to another with a grey level of 'j' under a predefined distance

and angle. From these matrices, sets of statistical measures are computed, called feature

vectors.

Energy: It is a gray-scale image texture measure of homogeneity changing, reflecting the

distribution of image gray-scale uniformity of weight and texture.

p(x,y) is the GLC M

Contrast: Contrast is the main diagonal near the moment of inertia, which measure the value

of the matrix is distributed and images of local changes in number, reflecting the image

clarity and texture of shadow depth.



Correlation Coefficient: Measures the joint probability occurrence of the specified pixel

pairs. Correlation: sum(sum((x- μx)(y-μy)p(x , y)/σxσy))

Homogeneity: Measures the closeness of the distribution of elements in the GLCM to the

GLCM diagonal.

Homogeneity = sum (sum (p(x, y)/ (1 + [x-y])))

Drawbacks

Poor discriminatory power

High computational load

Loss of edge details due to shift variant property.

5.2 KNN Classifier

In pattern recognition, the k-nearest neighbour algorithm (k-NN) is a method

for classifying objects based on closest training examples in the feature space. k-NN is a type

of instance-based learning, or lazy learning where the function is only approximated locally

and all computation is deferred until classification. The k-nearest neighbour algorithm is

amongst the simplest of all machine learning algorithms: an object is classified by a majority

vote of its neighbours, with the object being assigned to the class most common amongst

its k nearest neighbours (k is a positive integer, typically small). If k = 1, then the object is

simply assigned to the class of its nearest neighbour.

The same method can be used for regression, by simply assigning the property value for the

object to be the average of the values of its nearest neighbours. It can be useful to weight the

contributions of the neighbours, so that the nearer neighbours contribute more to the average

than the more distant ones. (A common weighting scheme is to give each neighbour a weight

of 1/d, where d is the distance to the neighbour. This scheme is a generalization of linear

interpolation.)

5.3 NEURAL NETWORK

Neural networks are predictive models loosely based on the action of biological neurons.

The selection of the name “neural network” was one of the great PR successes of the

Twentieth Century. It certainly sounds more exciting than a technical description such as “A



network of weighted, additive values with nonlinear transfer functions”. However, despite the

name, neural networks are far from “thinking machines” or “artificial brains”. A typical

artificial neural network might have a hundred neurons. In comparison, the human nervous

system is believed to have about 3x1010 neurons. We are still light years from “Data”.

The original “Perceptron” model was developed by Frank Rosenblatt in 1958.

Rosenblatt’s model consisted of three layers, (1) a “retina” that distributed inputs to the

second layer, (2) “association units” that combine the inputs with weights and trigger a

threshold step function which feeds to the output layer, (3) the output layer which combines

the values. Unfortunately, the use of a step function in the neurons made the perceptions

difficult or impossible to train. A critical analysis of perceptron’s published in 1969 by

Marvin Minsky and SeymorePapert pointed out a number of critical weaknesses of

perceptron’s, and, for a period of time, interest in perceptron’s waned.

Interest in neural networks was revived in 1986 when David Rumelhart, Geoffrey Hinton and

Ronald Williams published “Learning Internal Representations by Error Propagation”. They

proposed a multilayer neural network with nonlinear but differentiable transfer functions that

avoided the pitfalls of the original perceptron’s step functions. They also provided a

reasonably effective training algorithm for neural networks.

5.3.1 Types of Neural Networks:

1) Artificial Neural Network

2) Back propagation networks

3) General Regression Neural Networks

DTREG implements the most widely used types of neural networks:

a) Multilayer Perceptron Networks (also known as multilayer feed-forward network),

b) Back propagation networks (BPN)



(a)The Multilayer Perceptron Neural Network Model

The following diagram illustrates a perceptron network with three layers:

This network has an input layer (on the left) with three neurons, one hidden layer (in the

middle) with three neurons and an output layer (on the right) with three neurons.

There is one neuron in the input layer for each predictor variable. In the case of categorical

variables, N-1 neurons are used to represent the N categories of the variable.

Input Layer — A vector of predictor variable values (x1...xp) is presented to the input layer.

The input layer (or processing before the input layer) standardizes these values so that the

range of each variable is -1 to 1. The input layer distributes the values to each of the neurons

in the hidden layer. In addition to the predictor variables, there is a constant input of 1.0,

called the bias that is fed to each of the hidden layers; the bias is multiplied by a weight and

added to the sum going into the neuron.

Hidden Layer — Arriving at a neuron in the hidden layer, the value from each input neuron

is multiplied by a weight (wji), and the resulting weighted values are added together

producing a combined value uj. The weighted sum (uj) is fed into a transfer function, σ, which

outputs a value hj. The outputs from the hidden layer are distributed to the output layer.

Output Layer — Arriving at a neuron in the output layer, the value from each hidden layer

neuron is multiplied by a weight (wkj), and the resulting weighted values are added together

producing a combined value vj. The weighted sum (vj) is fed into a transfer function, σ, which

outputs a value yk. The y values are the outputs of the network.



If a regression analysis is being performed with a continuous target variable, then there is a

single neuron in the output layer, and it generates a single y value. For classification

problems with categorical target variables, there are N neurons in the output layer producing

N values, one for each of the N categories of the target variable.

(b)Back propagation networks (BPN):

Back Propagation (BPN) and General Regression Neural Networks (GRNN) have similar

architectures, but there is a fundamental difference: Probabilistic networks perform

classification where the target variable is categorical, whereas general regression neural

networks perform regression where the target variable is continuous. If you select a

BPN/GRNN network, DTREG will automatically select the correct type of network based on

the type of target variable.

Architecture of a BPN:

All BPN networks have four layers:

Input layer — There is one neuron in the input layer for each predictor variable. In the case

of categorical variables, N-1 neurons are used where N is the number of categories. The input

neurons (or processing before the input layer) standardize the range of the values by

subtracting the median and dividing by the interquartile range. The input neurons then feed

the values to each of the neurons in the hidden layer.

Hidden layer — This layer has one neuron for each case in the training data set. The neuron

stores the values of the predictor variables for the case along with the target value. When

presented with the x vector of input values from the input layer, a hidden neuron computes



the Euclidean distance of the test case from the neuron’s center point and then applies the

RBF kernel function using the sigma value(s). The resulting value is passed to the neurons in

the pattern layer.

Pattern layer / Summation layer — The next layer in the network is different for BPN

networks and for GRNN networks. For BPN networks there is one pattern neuron for each

category of the target variable. The actual target category of each training case is stored with

each hidden neuron; the weighted value coming out of a hidden neuron is fed only to the

pattern neuron that corresponds to the hidden neuron’s category. The pattern neurons add the

values for the class they represent (hence, it is a weighted vote for that category).

Decision layer — The decision layer is different for BPN and GRNN networks. For BPN

networks, the decision layer compares the weighted votes for each target category

accumulated in the pattern layer and uses the largest vote to predict the target category.

5.3.2 BACK PROPAGATION ALGORITHM

Consider a network with a single real input x and network function F. The

derivative F’(x) is computed in two phases:

Feed-forward: the input x is fed into the network. The primitive functions at the nodes and

their derivatives are evaluated at each node. The derivatives are stored.

Back propagation: The constant 1 is fed into the output unit and the network

is run backwards. Incoming information to a node is added and the result is multiplied by the

value stored in the left part of the unit. The result is transmitted to the left of the unit. The

result collected at the input unit is the derivative of the network function with respect to x.

STEPS OF THE ALGORITHM

The back propagation algorithm is used to compute the necessary corrections, after

choosing the weights of the network randomly. The algorithm can be decomposed in the

following four steps:

i) Feed-forward computation.

ii) Back propagation to the output layer.

iii) Back propagation to the hidden layer and Weight updates.



CHAPTER-6

RESULTS

MATLAB

MATLAB is a high-performance language for technical computing. It integrates

computation, visualization, and programming in an easy-to-use environment where problems

and solutions are expressed in familiar mathematical notation. Typical uses include

Typical uses of MATLAB

Math and computation

Algorithm development

Data acquisition

Modeling, simulation, and prototyping

Data analysis, exploration, and visualization

Scientific and engineering graphics

Application development, including graphical user interface MATLAB brings to

digital image processing is an extensive set of functions for processing multidimensional

arrays of which images (two-dimensional numerical arrays) are a special case. The Image

Processing Toolbox is a collection of functions that extend the capability of the MATLAB

numeric computing environment. These functions, and the expressiveness of the MATLAB

language, make image-processing operations easy to write in a compact, clear manner, thus

providing an ideal software prototyping environment for the solution of image processing

problems. In this chapter we introduce the basics of MATLAB notation, discuss a number of

fundamental toolbox properties and functions, and begin a discussion of programming

concepts. Thus, the material in this chapter is the foundation for most of the software-related

discussions in the remainder of the book.

An image may be defined as a two-dimensional function fxy (,), where x and y are

spatial (plane) coordinates, and the amplitude of f at any pair of coordinates is called the

intensity of the image at that point.



6.1 THE MAIN FEATURES OF MATLAB

Advance algorithm for high performance numerical computation, especially in the

Field matrix algebra a large collection of predefined mathematical functions and the ability to

define one’s own functions.

Two-and three-dimensional graphics for plotting and displaying data

A complete online help system

Powerful, matrix or vector oriented high-level programming language for individual

applications.

Toolboxes available for solving advanced problems in several application areas

Matrices in MATLAB are stored in variables with names such as A, a, RGB, real

array, and so on. Variables must begin with a letter and contain only letters, numerals, and

underscores. As noted in the previous paragraph, all MATLAB quantities in this book are

written using monospace characters. We use conventional Roman, italic notation, such as fxy

(,), for mathematical expressions. Images are read into the MATLAB environment using

function imread, whose basic syntax is imread ('filename') MATLAB documentation uses the

terms matrix and array interchangeably. However, keep in mind that a matrix is two

dimensional, whereas an array can have any finite dimension. Recall from Section 1.6 that we

use margin icon.

6.2 CAPABILITIES OF MATLAB

MATLAB is an interactive system whose basic data element is an array that does not

require dimensioning. This allows you to solve many technical computing problems,

especially those with matrix and vector formulations, in a fraction of the time it would take to

write a program in a scalar non interactive language such as C or FORTRAN.

The name MATLAB stands for matrix laboratory. MATLAB was originally written to

provide easy access to matrix software developed by the LINPACK and EISPACK projects.

Today, MATLAB engines incorporate the LAPACK and BLAS libraries, embedding

the state of the art in software for matrix computation.

MATLAB has evolved over a period of years with input from many users. In

university environments, it is the standard instructional tool for introductory and advanced



courses in mathematics, engineering, and science. In industry, MATLAB is the tool of choice

for high productivity research, development, and analysis.

MATLAB features a family of add-on application-specific solutions called toolboxes.

Very important to most users of MATLAB, toolboxes allow you to learn and apply

specialized technology. Toolboxes are comprehensive collections of MATLAB functions (M-

files) that extend the MATLAB environment to solve particular classes of problems. Areas in

which toolboxes are available include signal processing, control systems, neural networks,

fuzzy logic, wavelets, simulation, and many others. MATLAB features a family of add-on

application specific solutions called toolboxes. Very important to most users of MATLAB,

toolboxes allow you to learn and apply specialized technology. Toolboxes are comprehensive

collections of MATLAB functions (M-files) that extend the MATLAB environment to solve

particular classes of problems. Areas in which toolboxes are available include signal

processing, control systems, neural networks, fuzzy logic, wavelets, simulation, and many

others.

6.3 MATLAB SYSTEM

The MATLAB system consists of five main parts:

6.3.1 Development Environment

This is the set of tools and facilities that help you use MATLAB functions and files.

Many of these tools are graphical user interfaces. It includes the MATLAB desktop and

Command Window, a command history, an editor and debugger, and browsers for viewing

help, the workspace, files, and the search path.

6.3.2 The MATLAB Mathematical Function

This is a vast collection of computational algorithms ranging from elementary

functions like sum, sine, cosine, and complex arithmetic, to more sophisticated functions like

matrix inverse, matrix Eigen values, Bessel functions, and fast Fourier transforms.

6.3.3 The MATLAB Language

This is a high-level matrix/array language with control flow statements, functions,

data structures, input/output, and object-oriented programming features. It allows both

"programming in the small” to rapidly create quick and dirty throw-away programs, and



"programming in the large" to create complete large and complex application programs.

6.3.4 Graphics

MATLAB has extensive facilities for displaying vectors and matrices as graphs, as

well as annotating and printing these graphs. It includes high-level functions for two-

dimensional and three-dimensional data visualization, image processing, animation, and

presentation graphics. It also includes low-level functions that allow you to fully customize

the appearance of graphics as well as to build complete graphical user interfaces on your

MATLAB applications.

6.3.5 The MATLAB Application Program Interface (API)

This is a library that allows you to write C and FORTRAN programs that interact with

MATLAB. It includes facilities for calling routines from MATLAB (dynamic linking),

calling MATLAB as a computational engine, and for reading and writing MAT-files.

6.4 MATLAB WORKING ENVIRONMENT

6.4.1 MATLAB DESKTOP

Fig 6.1 MATLAB Command window

MATLAB Desktop is the main MATLAB application window. The desktop contains five sub

windows, the command window, the workspace browser, the current directory window, the



command history window, and one or more figure windows, which are shown only when the

user displays a graphic. This is a high-level matrix/array language with control flow

statements, functions, data structures, input/output, and object oriented programming

features. It allows both "programming in the small" to rapidly create quick and dirty throw-

away programs, and "programming in the large" to create complete large and complex

application programs. This is a high-level matrix/array language with control flow

statements, functions, data structures, input/output, and object-oriented programming

features. It allows both "programming in the small" to rapidly create quick and dirty throw-

away programs, and "programming in the large" to create complete large and complex

application programs. The current Directory tab above the workspace tab shows the contents

of the current directory, whose path is shown in the current directory window For example, in

the windows operating system the path might be as follows: C:\MATLAB\Work, indicating

that directory “work” is a subdirectory of the main directory “MATLAB”.

6.5 RESULTS

Fig 6.1 A CT scanned image is given to the GUI



Fig6.2 An image is converted to Discrete Wavelet Transform image

Fig 6.3 Calculating percentage of cancer and detecting cancer stage by loading the database



Fig 3.4 Calculation of cancer effected area



CHAPTER-7

CONCLUSION

Cancer is potentially fatal disease. Detecting cancer is still challenging for the doctors in the

field of medicine. Even now the actual reason and complete cure of cancer is not invented.

Detection of cancer in earlier stage is curable. In this work we have developed a system

called image processing-based cancer prediction system. The main aim of this model is to

provide the earlier warning to the users and it is also cost and time saving benefit to the user.

Edges in image processing help us to determine objects. In this method we have successfully

identified the cancerous nodules in the lung by using their CT scan images. Physicians use

the naked eye to detect the growth and spread of cancerous nodule in the lungs from the CT

scan images. This may incorporate human error in detection and it is quite tedious too. The

method we propose automatically detects and identifies the cancerous cells of the lungs. It

can also help in determining the shape and pattern of the nodule which provide necessary

information needed for the proper medication and also helps in determining the area affected

by the cancerous cells. It also identifies the cells which might have been unnoticed by human

eyes. Detection of lung cancer at an early stage can be difficult but using the proposed work

detection becomes uncomplicated and the chances of the early treatment of the patient and

therefore chances of survival of the patient increases. Thus, using an automated system not

only reduces chances of human error but also increases the accuracy up to 90%.



REFERENCES

[1] Matsumoto M, Horikoshi H, Moteki T, et al. A pilot study with lungcancer screening CT

(LSCT) at the secondary screening for lung cancer detection. Nippon Acta Radiol 1995;

55:172-17.

[2] Kaneko M, Eguchi K, Ohmatsu H, et al. Peripheral lung cancer: screening and detection

with low-dose spiral CT versus radiography. Radiology 1996; 201:798-802.

[3] P Lillington GA. Management of solitary pulmonary nodules: how to decide when

resection is required. Postgrad Med 1997; 101(3): 145- 50.

[4] L.E Romans, Introduction to computed tomography, springer (1995)

[5] R. G. Gonzale and R.E. Woods, Digital Image processing, Addison Wesley, New York

(1993).

[6] Weixing. Wang, 2005,A segment algorithm for rock fractures detection, International

conference on Advances in Pattern Recognition 3th (ICAPPR 2005), in Bath, UK 22-25 of

August 2005,pp.580-588.

[7] Marr D. and E. Hildreth, Theory of Edge detection, Proc. Royal Society of London, vol.

B-207, 1980, pp. 187-217.

[8] Canny J., A computational Approach to Edge Detection, IEEE Trans. Pattern Analysis

and Machine Intelligence, vol. 8, no. 6, 1986, pp. 679-698.3.,van Leeuwen, J. (ed.):

Computer Science Today. Recent Trends and Developments. Lecture Notes in Computer

Science, Vol. 1000. Springer-Verlag, Berlin Heidelberg New York (1995).

[9] Rubins JB, Rubins HB, Temporal trends in the prevalence of malignancy in resected

solitary pulmonary lesions. Chest 1996; 109(1): 100-103..

[10] Shaham D, Guralnik L, The solitary pulmonary nodule: radiologic considerations. Semin

Ultrasound CT MR 2000; 21(2): 97-115.

[11] ZHAO.S and WANG .G: Wavelet operator and their applications in computerized

tomography, proc, SPIE, 997, pp: 337-348.

[12] Quint LE, Park CH, lannettoni MD. Solitary pulmonary nodules in patients with

extrapulmonary neoplasms. Radiology 2000; 217(1): 257-606



[13] Heath M.D., S. Sarkar, T. Sanocki, and K.W. Bowyer, A Robust Visual Method for

Assessing the Relative Performance of Edge-Detection Algorithms, IEEE Trans. Pattern

Analysis and Machine Intelligence, vol. 19, no. 12, 1997, pp. 1338-1359.

[14] Swensen SJ, Viggiano RW, Midthun DE, et al. Lung nodule enhancement at CT:

multicenter study. Radiology 2000; 214(1): 73- 80.

[15] Bergholm F., Edge focusing, IEEE Trans. Pattern Analysis and Machine Intelligence,

vol. 9, 1987, pp. 726-741.

[16] Goldsmith SJ, Kostakoglu L. Role of nuclear medicine in the evaluation of the solitary

pulmonary nodule. Semin Ultrasound CT MR 2000; 21(2): 129-138.



APPENDIX

Main code

function varargout = gui(varargin)% GUI M-file for gui.fig% GUI, by itself, creates a new GUI or raises the existing% singleton*.%% H = GUI returns the handle to a new GUI or the handle to% the existing singleton*.%% GUI('CALLBACK',hObject,eventData,handles,...) calls the local% function named CALLBACK in GUI.M with the given input arguments.%% GUI('Property','Value',...) creates a new GUI or raises the% existing singleton*. Starting from the left, property value pairs are% applied to the GUI before gui_OpeningFcn gets called. An% unrecognized property name or invalid value makes property application% stop. All inputs are passed to gui_OpeningFcn via varargin.%% *See GUI Options on GUIDE's Tools menu. Choose "GUI allows only one% instance to run (singleton)".%% See also: GUIDE, GUIDATA, GUIHANDLES

% Edit the above text to modify the response to help gui

% Last Modified by GUIDE v2.5 16-Feb-2015 17:08:09

% Begin initialization code - DO NOT EDITgui_Singleton = 1;gui_State = struct('gui_Name', mfilename, ... 'gui_Singleton', gui_Singleton, ... 'gui_OpeningFcn', @gui_OpeningFcn, ... 'gui_OutputFcn', @gui_OutputFcn, ... 'gui_LayoutFcn', [] , ... 'gui_Callback', []);if nargin && ischar(varargin{1}) gui_State.gui_Callback = str2func(varargin{1});end

if nargout [varargout{1:nargout}] = gui_mainfcn(gui_State, varargin{:});else gui_mainfcn(gui_State, varargin{:});end% End initialization code - DO NOT EDIT



% --- Executes just before gui is made visible.function gui_OpeningFcn(hObject, eventdata, handles, varargin)% This function has no output args, see OutputFcn.% hObject handle to figure% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)% varargin command line arguments to gui (see VARARGIN)

% Choose default command line output for guihandles.output = hObject;a = ones(256,256);axes(handles.axes1);imshow(a);axes(handles.axes2);imshow(a);axes(handles.axes4);imshow(a);set(handles.text1,'string','');% Update handles structureguidata(hObject, handles);

% UIWAIT makes gui wait for user response (see UIRESUME)% uiwait(handles.figure1);

% --- Outputs from this function are returned to the command line.function varargout = gui_OutputFcn(hObject, eventdata, handles) % varargout cell array for returning output args (see VARARGOUT);% hObject handle to figure% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

% Get default command line output from handles structurevarargout{1} = handles.output;

% --- Executes on button press in Browse_im.function Browse_im_Callback(hObject, eventdata, handles)% hObject handle to Browse_im (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

cd timages [file,path] = uigetfile('*.jpg;*.bmp;*.gif;*.png', 'Pick an Image File'); %% Image selection process im = imread(file); cd .. im=imresize(im,[256 256]);



if size(im,3)>1 im = rgb2gray(im); end figure; imshow(im); title('Input Image'); mfima=medfilt2(im,[3 3]); axes(handles.axes1); imshow(mfima); title('Medianfilter Image');handles.im = im;handles.mfima =mfima;

% Update handles structureguidata(hObject, handles);% helpdlg('Test Image Selected');

% --- Executes on button press in database_load.function database_load_Callback(hObject, eventdata, handles)% hObject handle to database_load (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

nnlearn;

% --- Executes on button press in classify_im.function classify_im_Callback(hObject, eventdata, handles)% hObject handle to classify_im (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

%%%%%Importing the trained network parametersload qfeat;load netp;%%%%%%classification

cout = sim(netp,qfeat); cout = vec2ind(cout);

if isequal(cout,1) set(handles.text6,'String','Stage :'); set(handles.text1,'String',' 0% Effected [Normal]'); elseif isequal(cout,2) set(handles.text6,'String','Stage :');



set(handles.text1,'String','Cancer 30% Effected [BENIGN]'); elseif isequal(cout,3) set(handles.text6,'String','Stage :'); set(handles.text1,'String','Cancer 50% Effected [MALIGNANT]'); else helpdlg('Db updation required');

end handles.result = cout; guidata(hObject,handles); % --- Executes on button press in transform.function transform_Callback(hObject, eventdata, handles)% hObject handle to transform (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

im = handles.im;

[LL LH HL HH] = dwt2(im,'db1'); %% HAAR,DB,Bi.ortho

aa = [LL LH;HL HH];

% % % % 2nd level decomp[LL1 LH1 HL1 HH1] = dwt2(LL,'db1');

% aa1 = [LL1 LH1;HL1 HH1];

% % % 3rd level Decomp

[LL2 LH2 HL2 HH2] = dwt2(LL1,'db1');

% % % 4th level Decomp

[LL3 LH3 HL3 HH3] = dwt2(LL2,'db1');

aa1 = [LL3 LH3;HL3 HH3];

aa2 = [aa1 LH2;HL2 HH2];

aa3 = [aa2 LH1;HL1 HH1]; aa4 = [aa3 LH;HL HH];



axes(handles.axes2);imshow(aa2,[]);title('Discrete Wavelet Transform Image');

% % % Select the wavelet coefficients LH3 and HL3% % % Haralick features for LH3

LH3 = uint8(LH3);Min_val = min(min(LH3));Max_val = max(max(LH3));level = round(Max_val - Min_val);GLCM = graycomatrix(LH3,'GrayLimits',[Min_val Max_val],'NumLevels',level);stat_feature = graycoprops(GLCM);Energy_fet1 = stat_feature.Energy;Contr_fet1 = stat_feature.Contrast;Corrla_fet1 = stat_feature.Correlation;Homogen_fet1 = stat_feature.Homogeneity;

% % % % % Entropy R = sum(sum(GLCM)); Norm_GLCM_region = GLCM/R; Ent_int = 0; for k = 1:length(GLCM)^2 if Norm_GLCM_region(k)~=0 Ent_int = Ent_int + Norm_GLCM_region(k)*log2(Norm_GLCM_region(k)); end end Entropy_fet1 = -Ent_int;

%%%%%Haralick Features For HL3 HL3 = uint8(HL3);Min_val = min(min(HL3));Max_val = max(max(HL3));level = round(Max_val - Min_val);GLCM = graycomatrix(HL3,'GrayLimits',[Min_val Max_val],'NumLevels',level);stat_feature = graycoprops(GLCM);Energy_fet2 = stat_feature.Energy;Contr_fet2 = stat_feature.Contrast;Corrla_fet2= stat_feature.Correlation;Homogen_fet2 = stat_feature.Homogeneity;% % % % % Entropy R = sum(sum(GLCM)); Norm_GLCM_region = GLCM/R; Ent_int = 0; for k = 1:length(GLCM)^2 if Norm_GLCM_region(k)~=0 Ent_int = Ent_int + Norm_GLCM_region(k)*log2(Norm_GLCM_region(k)); end



end% % % % % % Ent_int = entropy(GLCM); Entropy_fet2 = -Ent_int;

%%%%% Feature Sets

F1 = [Energy_fet1 Contr_fet1 Corrla_fet1 Homogen_fet1 Entropy_fet1];F2 = [Energy_fet2 Contr_fet2 Corrla_fet2 Homogen_fet2 Entropy_fet2];

qfeat = [F1 F2]';save qfeat qfeat;

disp('Query Features: ');disp(qfeat);

% --- Executes on button press in close.function close_Callback(hObject, eventdata, handles)% hObject handle to close (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

delete *.mat;close all;

% --- Executes on button press in clear.function clear_Callback(hObject, eventdata, handles)% hObject handle to clear (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)clc;set(handles.text1,'string','');set(handles.text6,'string','');set(handles.text7,'string','');set(handles.text8,'string','');set(handles.text9,'string','');

a = ones(256,256);axes(handles.axes1);imshow(a);axes(handles.axes2);imshow(a);

clear all;

% --- Executes on button press in validate.function validate_Callback(hObject, eventdata, handles)% hObject handle to validate (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB



% handles structure with handles and user data (see GUIDATA)

%%%%Parameters Evaluation %%%%%%total number of test samples 9 Tp = 3; Fn = 2; %%%%%%%after classification Fp = 2; Tn = 3; %%%%%Tp --> Abnormality correctly classified as abnormal %%%%%Fn --> Abnormality incorrectly classified as normal %%%%%Fp --> Normal incorrectly classified as abnormal %%%%%Tn --> Normal correctly classified as normal Sensitivity = (Tp./(Tp+Fn)).*100;Specificity = (Tn./(Tn+Fp)).*100;

Accuracy = ((Tp+Tn)./(Tp+Tn+Fp+Fn)).*100;

figure('Name','Performance Metrics','MenuBar','none'); bar3(1,Sensitivity,0.3,'m');hold on;bar3(2,Specificity,0.3,'r');hold on;bar3(3,Accuracy,0.3,'g');hold off;

xlabel('Parametrics--->');zlabel('Value--->');legend('Sensitivity','Specificity','Accuracy');

disp('Sensitivity: '); disp(Sensitivity);disp('Specificity: '); disp(Specificity);disp('Accuracy:'); disp(Accuracy);

% --- Executes on button press in segment.function segment_Callback(hObject, eventdata, handles)% hObject handle to segment (see GCBO)% eventdata reserved - to be defined in a future version of MATLAB% handles structure with handles and user data (see GUIDATA)

result =handles.result;

%% In CLassification Condition we got result for cout=1/2/3...%%Depend upon this result our input image is Stage 1 or 2 or 3

inp = handles.im;

if result ==1 warndlg ('No Tumor'); elseif result ==2



[segout,tarea] = BSegment(inp); boundary = bwboundaries(im2bw(segout)); axes(handles.axes1); imshow(inp); title('Tumor Area Localization'); hold on; for ii=1:1:length(boundary) btemp = boundary{ii}; plot(btemp(:,2),btemp(:,1),'r','LineWidth',4); end hold off; axes(handles.axes4); imshow(segout); title('Segmented Image'); set(handles.text7,'String','Area :'); set(handles.text8,'String',tarea); set(handles.text9,'String','mm.^2'); handles.area = tarea; guidata(hObject,handles); elseif result ==3 [segout,tarea] = MSegment(inp); boundary = bwboundaries(im2bw(segout)); axes(handles.axes1); imshow(inp); title('Tumor Area Localization'); hold on; for ii=1:1:length(boundary) btemp = boundary{ii}; plot(btemp(:,2),btemp(:,1),'r','LineWidth',4); end hold off; axes(handles.axes4); imshow(segout); title('Segmented Image'); set(handles.text7,'String','Area :'); set(handles.text8,'String',tarea); set(handles.text9,'String','mm.^2'); handles.area = tarea; guidata(hObject,handles); end




Date post:	02-Aug-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

LUNG CANCER DETECTION ON CT IMAGES BY USING WAVELET … · Web view2019-07-21 · Clustering...

Documents