PEDIATRIC MEASUREMENT DEVICE FOR EARLY
DETECTION OF PLAGIOCEPHALY
IPHONE APP EDITIONCODY HALL
CHRIS HEELANJORGE PEREZ
CRANIOMETERProject goal
• To develop a low cost, safe, and accurate plagiocephaly measurement device for children aged infant to 24 months
• Digital solution that limits complexity as to lower price
Initial design• Physical Device
• Two digitial calipers used to detect linear distances and angles
CRANIOMETER PROS/CONS
PROSSimple Theory
• Directly sense angle using sensor• Calculate distances using
trigonometry
Simple UI• Three push buttons• Results clearly displayed on LCD
Easier to use with squirming childSimple design
• Increased reliability
Requires small investment by pediatrician
CONSHardware design
• PCB design/debugging• Component selection
Mechanical design• AutoCAD parts
Purchasing physical partsAssembling physical partsRequires physical contact with the infantThird-party required for manufacturing/marketing
IPHONE APPLICATIONNew design
• iPhone application• Use camera to take image of
infant’s head• Markers on paper provide
absolute distance calibration• PDF included with app
• Perform image analysis to calculate angles/linear distances
• Display results on phone
IPHONE APPLICATION PROS/CONS PROSUser already owns physical device
• No hardware to manufacture/purchase
Safer• No physical contact with child
App Store provides simple distribution methodFocus on purely software will allow for a more polished and professional productPotential Market
• Estimated 58,000 pediatricians in the United Stated (American Academy of Pediatrics)
• 94% of doctors own smartphones (compared to 59% in 2006)
• Doctors prefer iPhone (44%) to Blackberry (25%) and Andriod (25%)
• Estimated 24,000 potential customers• Verizon adding iPhone this year
CONSMore complex theory
• Calibrate pixels to distance using known markers in image
• Edge-detection processing to calculate linear distances
Requires pediatrician to own iPhoneMore complicated UI
• Also more complicated to design
Harder to use with squirming child• Camera must be parallel to child’s head
Requires designers to learn Cocoa (Objective-C) programming
HOW DO WE KNOW THIS WILL WORK?iPhone 4 Camera Specifications
• Back- Illuminated CMOS Sensor• 5-Megapixel Resolution (2592 x
1936)• Fixed aperture f/2.8• Permanent 3.85mm focal length
IPHONE CAMERA TESTINGThird party research
• <0.2% barrel distortion = Very GoodLow noise
• <2% under fluorescent lighting (ISO 144)Need to determine sharpness
• Rise distance
RESOLUTION30cm x 22cm area / 2592 x 1936 pixelsGives us a 0.13mm spatial resolution under perfect conditions
SEGMENTATION AND EDGE DETECTIONUse of histogram threshold
CONNECTED COMPONENT LABELING
NECESSARY LABELING METHOD TO MARK DIFFERENT CIRCLES
FIND THE CENTERS OF CIRCLES- STRIKE A CHORD
Find the maximum distance from one point on the edge to anotherTake midpoint as centerDistance between centers is determines pixel size
EDGE DETECTION ON A REAL PHOTO
CHALLENGESDeveloping efficient algorithms to work on iPhoneAutomated or assisted point selection process- GUI controlMaintaining millimeter resolution during heavy image processing
IPHONE APP DEVELOPMENTRequired
Computer running Mac OS-two designers own OSX machines
Register with Apple as a developer-two designers registered
Download/install SDK-two designers have successfully installed/used the SDK
Create application
Submit application to App Store for approval
HELLO WORLD! APP
APPLICATION PROGRAMMING
Learning Cocoa (Objective-C)• Stanford iPhone app development
lectures• Free from iTunes University• Full semester course (lecture videos
and PPT slides)• Other online coding/app tutorials• One team member proficient in C
programming• Layer underneath Objective-C
• One team member proficient in Java programming
• Object-oriented programming
IMAGE PROCESSING ALGORITHMSImageJ
• Free image processing program• Open source Java code• Port needed algorithms to Objective-C
• One designer proficient in JavaIndependent research
• Image processing tutorials online in many different coding languages
IMAGE PROCESSING LOCATIONTwo choices for image processing
• All processing written within application (Objective-C)• App is completely localized
• Processing performed on server running Python script• App captures/displays image and provides GUI• App sends image to server• Image processing offloaded to Python• Script sends results back to app