Methodology and Test Results

The malignant skin disease detection method implemented in "ProRodinki" App is based on the analysis of smartphone-taken photographs and risk-factors with a trained Neural Network. The method was proposed by the head of the Department of Skin and Venereal Diseases of the “PIMU” Medical Research University, Professor Shlivko.  

 

The method is based on many years of experience in treating patients with malignant skin diseases at the University Clinic. The following clinical signs of skin “spots” help a doctor to diagnose skin cancer:

  • Colour

  • Size

  • Ulceration

  • Borders

  • Evolution

  • Localization 

  • Other additional signs

    

More experienced doctors deliver more accurate diagnosis at earlier stages as the long-term observations indicate. The availability of the experienced doctors is limited, patients are routed to them in already suspected cases. The early detection becomes even more important with the increasing skin cancer cases worldwide.

 

The idea was to transfer the accumulated experience of the best doctors to Artificial Intelligence by training a Neural Network on archives with several thousand verified cases.

 

The methodology was implemented in collaboration with a group of highly qualified dermatologists and oncologists from PIMU under Prof. Shilvko leadership. The methodology was delivered by Information Technology (IT) and Artificial Intelligence (AI) experts from AIMED LLC. 

Prorodinki App was tested by May 2020 on more than 1,000 suspected skin disease cases, including hemangiomas, nevi, seborrheic keratosis, melanomas, basal cell carcinoma and others. Each recorded image was independently evaluated by at least two PIMU experts.

The App demonstrated characteristics on par with an experienced dermatologist:

  • Sensitivity (the ability of a test to correctly identify patients with a disease): 88%

  • Specificity (the ability of a test to correctly identify people without the disease): 78%

  • Accuracy: 81%

DISCLAIMER EXCERPTS (Please, follow the link for the complete disclaimer):

  • Prorodinki App cannot be used for the treatment and diagnosis of diseases

  • The complex issues only a recommendation about visiting a doctor and choosing a doctor of the desired specialization based on the analysis of the sent materials

  • When receiving recommendations, it is necessary to keep in mind their probabilistic nature in accordance with the values ​​of the characteristics given above.