Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
1 Detection of skin cancer
Detection of skin cancer
isp.imm.dtu.dk
Jan Larsen
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
2 Detection of skin cancer
ISP Group
HumanitarianDemining
Monitor Systems
Biomedical
Neuroinformatics
Multimedia
Machinelearning
•3+1 faculty•6+1 postdocs•20 Ph.D. students•10 M.Sc. students
•3+1 faculty•6+1 postdocs•20 Ph.D. students•10 M.Sc. students
from processing to understanding
extraction of meaningful information by learning
Biomedical
•Neuroimaging(PET,EEG,fMRI)
•EEG sensor for early warning of low blood suguar
•Improved SP in hearing aids
10 phd students, 3 post docs
www.intelligentsound.org
www.cimbi.org
hendrix.imm.dtu.dk
isp.imm.dtu.dk
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
3 Detection of skin cancer
Skin cancer •More than 800 cases in Denmark yearly
•Annual increase 5-10%
•Benign nevi
•Atypical nevi
•Malignant melanoma
•Inexperienced doctors detect 31%
•Experienced doctors detect 63-75%
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
4 Detection of skin cancer
Objectives
Develop a cost-effective and practical tool for diagnosis supportGain more insight into the understanding of factors in the development of skin cancer
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
5 Detection of skin cancer
Cross-disciplinary research
Signal and Image processing
StatisticsMachine learning
Domain knowledge
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
6 Detection of skin cancer
Outline
Machine learning framework for skin cancer detection– Involves all issues of machine learning
An image processing system for skin cancer detection– Involves feature selection, projection and integration– Involves linear and nonlinear classifiers
Other approachesSummary
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
7 Detection of skin cancer
The potential of learning machines
Most real world problems are too complex to be handled by classical physical modelsIn most real world situations there is access to data describing properties of the problemLearning machines can offer– Learning of optimal prediction/decision/action– Adaptation to the usage environment– New insights into the problem and suggestions for
improvement
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
8 Detection of skin cancer
A short history of learning machines
clas
sica
l
moder
n
ADALINE
Neural nets
Gaussian processes
Kernel machines
Mixture of experts
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
9 Detection of skin cancer
Issues in machine learning
Data
•quantity
•stationarity
•quality
•structure
Features
•representation
•selection
•extraction
•integration
Models
•structure
•type
•learning
•selection and
integration
•unsupervised
•semi-supervised
•supervised
•cost function
•maximum likelihood
•Bayesian
•online vs. off-line
Evaluation
•performance
•robustness
•complexity
•interpretation and visualization
•HCI
•parametric: linear, nonlinear, mixture models
•non-parametric: kernel, Gaussian processes, clustering
•noise models
•integration of prior and domain knowledge
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
10 Detection of skin cancer
Dermatoscopy imaging technique
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
11 Detection of skin cancer
Domain knowledge – dematoscopic features
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
12 Detection of skin cancer
Feature extraction
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
13 Detection of skin cancer
Median filtering
Removal of impulsive noise
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
14 Detection of skin cancer
Feature extraction
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
15 Detection of skin cancer
Segmentation
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
16 Detection of skin cancer
Feature extraction
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
17 Detection of skin cancer
Assymetry
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
18 Detection of skin cancer
Feature extraction
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
19 Detection of skin cancer
Edge abruptness
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
20 Detection of skin cancer
Feature extraction
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
21 Detection of skin cancer
Color prototypes
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
22 Detection of skin cancer
Segmentation into color prototypes
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
23 Detection of skin cancer
Bayes classifier
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
24 Detection of skin cancer
Bayes classifier
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
25 Detection of skin cancer
Neural network classifier
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
26 Detection of skin cancer
Likelihood learning
Training set: N samples of related x(k) and classes y(k)
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
27 Detection of skin cancer
Generalization
How well are we doing on future data from the same problem?
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
28 Detection of skin cancer
Bias Variance dilemma
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
29 Detection of skin cancer
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
30 Detection of skin cancer
Confusion matrix
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
31 Detection of skin cancer
Other techniques – Raman spectroscopy
A NIR laser beam excites molecules in the skinThe Raman scattering is a frequency shift in the reflected light which is related to the molecule structure
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
32 Detection of skin cancer
Raman spectrum
•MM: malignant melanoma
•NV: pigmented navi
•BCC: basal cell carcinoma
•SK: seborrhoeickeratosis
•NOR: normal
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
33 Detection of skin cancer
Raman classification results
Ref: Sigurdur Sigurdsson *’s are predicted values using a NN
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
34 Detection of skin cancer
Further reading
Hintz-Madsen, M., A probabilistic framework for classification of dermatoscopic images, pp. 156, Informatics and Mathematical Modelling, Technical University of Denmark, DTU, 1998Sigurdsson, S., A Probabilistic Framework for Detection of Skin Cancer by Raman Spectra, pp. 202, Informatics and Mathematical Modelling, Technical University of Denmark, DTU, 2003 Have, A. S., Datamining on distributed medical databases, Informatics and Mathematical Modelling, Technical University of Denmark, DTU, 2003 Papers accessible via http://isp.imm.dtu.dk
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
35 Detection of skin cancer
Related courses
02451 Digital Signal Processing02457 Nonlinear Signal Processing02459 Machine Learning for Signal Processing02501 Digital image analysis, vision and computer graphics 02505 Medical Image Analysis 31565 Advanced topics in Biomedical Signal Processing
Intelligent Signal Processing Group, IMM, DTU / Jan Larsen
36 Detection of skin cancer
Summary
Machine learning is, and will become, an important component in most real world applicationsDesigning a system involves cross-disciplinary competence – domain knowledge, features, classifiers etc.Automatic detection of skin cancer for diagnosis support is possible