Skip to content

introductory project into computer vision. using classical classifiers for automatic disease detection for medical treatment

License

Notifications You must be signed in to change notification settings

mikasenghaas/skin-lesion-detection

Repository files navigation

Project 3: Image Recognition in Medical Treatment

Can we help diagnosing dangerous skin lesions using machine-learning based image recognition?


Group 9: Aidan Stocks, Nicola Clark, Hugo Reinicke, Jonas-Mika Senghaas

Project Description

A dermatologist in asked us to investigate whether some characteristics of skin lesions could be reliably measured with a smartphone app. The characteristics the dermatologist was especially interested in are: asymmetry, border and color. Our goal was first to measure two or more of these characteristics in a set of skin lesion images. You can choose which one(s) you want to implement, and you may also implement other features, if you want. Then you should try to assess how good the measurements are, for example, by predicting the diagnosis of the skin lesions based on these features.

Background and Motivation

The amount of medical imaging - just as data in any other field - has increased tremendously within the last decade, making it more and more difficult to manually inspect medical images for a diagnosis.

Furthermore, people have proven to be hesitant of visiting doctors because of seemingly 'light' changes in their skin, which did not seem to be important enough to occupy a doctor's time. With skin diseases being especially effective in treatment if detected early, this is fatal. An easy-to-use app that implements automated detection of skin diseases from the sofa, this issue would be solved, which would ultimately safe lives.

Dependencies

Here we list the packages used in the project

Some Visualisations of the Project

About

introductory project into computer vision. using classical classifiers for automatic disease detection for medical treatment

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published