Scientists “taught” a computer to “recognize” melanoma from photographs

A new diagnostic method will allow in the future to determine the disease using a mobile phone

Employees and students of the National Research Nuclear University MEPhI have learned to diagnose skin melanoma, a dangerous oncological disease resembling a large reddish mole, using digital photographs.

As MK was told at MEPhI, the work was carried out with the participation of doctors from the Central State Medical Academy of the Administration of the President of the Russian Federation and the Medical Institute of the Peoples’ Friendship University of Russia.

Since a malignant tumor – melanoma – develops from melanocytes – pigment cells that produce melanin, the developers of the new method took as its basis the analysis of individual small structural elements of the image of a large “mole” – lines, lumps, dots, circles, etc. Digital detailed photographs of melanoma for this, they are introduced into a special, pre-trained program, which issues a “verdict”. She was “trained” on 3,000 reference photographs showing lines, dots and lumps of normal moles and malignant tumors. Image analysis is based on the most advanced Kittler method in modern oncodermatology, which involves the assessment of the most informative characteristics of the object: color, asymmetry, texture, shape.

One of the authors of the invention, head of the Department of Computer Medical Systems Valentin Nikitaev, expressed confidence that as the system improves, it will be possible to diagnose melanoma using digital photographs using a mobile phone. Although, of course, the final decision will remain with the doctor.

The new diagnostic method is registered with the Federal Institute of Industrial Property.