Machine Learning in the Health Sector

Artificial intelligence can also be of great help to doctors in diagnostic imaging. For example, the system is capable of accurately grouping various skin lesions and contributes to the elimination of misdiagnosis.

Detection of Skin Lesions with Machine Vision

  • Base model: VGG–16 Kaggle modell
  • Dataset: HAM10000 – actinic keratosis, basal-cell carcinoma, benign keratosis, dermatofibroma, melanoma, melanocytic nevus, vascular skin lesion
  • Input: 224×224
  • Output: percentage probability of 6 possible skin lesions
Dermatofibroma: 0.998905
Melanoma: 0.969042

The following VGG-16 model that was trained with physician-controlled images can identify 6 types of skin lesions with 85% accuracy.

The AI is loading...

    Please Note: This is a prototype.

    Covid-19 patients x-ray analising with machine learning models (example, it does NOT reflect reality)

    Development: Detection of COVID-19 Cases from Chest X-rays

    Effective screening of infected patients is key in the fight against COVID-19. Early examinations revealed that chest X-rays showed abnormalities if infection was present. COVID-Net, is a community-developed software designed to diagnose COVID-19 infection based on chest X-rays (CXRs). It is open source and available to the public.

    A data set of 13,975 CXR images is currently available for development.

    Development can be accelerated with the expansion of the research and data scientist community. The result will be a very accurate and practical, deep learning based solution by which COVID-19 cases can be easily detected.

    More information on the project:  github.com/lindawangg/COVID-Net

    Covid Inference Result
    Prediction: COVID-19
    Normal: 0.031
    Pneumonia: 0.189
    COVID-19: 0.780

    Covid Inference Severity
    Geographic severity: 0.519
    Geographic extent score for right + left lung (0 - 8): 4.155
    Opacity severity: 0.388
    Opacity extent score for right + left lung (0 - 6): 2.329