DermalE-HealthRecognitionSkinner

An eHealth App Prototype For Skin Problems Diagnosis | Recommendation Systems, Neural Networks -Based Solution

The Result

A completed prototype for eHealth consumer app that provides preliminary skin problems diagnosis and recommendations based on image recognition analysis of the problematic skin area.


 

The Challenge

Mellivora Software’s founder has come up with an idea of an eHealth app that provides primary diagnosis for skin disorders without having to visit a doctor. The idea had to be technically validated by building a demo prototype for the above described challenge.


The Solution

The skin parameters were tuned for white skin type due to available photo samples used for the prototype. These parameters could easily be customized for other skin types:

Currently the program consists of few major parts. The first part distinguishes skin areas from other elements (hair, clothes, etc) on the photo:

 

The second part marks skin problem areas, which contain the most typical elements related to a certain disease:

The third part ranks this particular skin disorder type to a certain disease class:

In the consumer version of mobile app the system is meant to provide user with the decease class (or, in lay terms, primary diagnosis) that has the highest probability rate.

 

Check out the video demonstrating the Skinner app idea and how it works:

 


Technology Stack

  • CNN (convolutional neural network)
  • Library: Python (numpy)
  • Tensorflow

About Mellivora Software

Mellivora Software helps SME and Enterprise businesses build custom IT solutions for a bunch of specific industries. Our core expertise is focused around:

  • Big Data, DevOps and AWS technologies
  • NLP technologies
  • Data Science/Machine Learning