METHOD FOR DIAGNOSING SKIN MELANOMA Russian patent published in 2022 - IPC A61B5/00 G06T7/00 

FIELD: medicine.

SUBSTANCE: invention relates to medicine, namely, to oncology, and can be used to diagnose skin melanoma. Method includes the stages of recording a dermatoscopic video image of the skin; measuring and selecting the affected region in the video image; measuring the coordinates and segmenting the image of the characteristic area; selecting a diagnostic property of the image in order to form a diagnosis; measuring the parameters of said property and comparing the parameters of the selected property with a library of predetermined parameters of the property; forming a diagnosis based on the comparison of the parameters of the property selected from the following group of properties: colour, border, asymmetry, and texture of the image, to the parameters of the diagnostic property of the image. After the dermatoscopic video image of the skin region is recorded and before the diagnostic property of the image for the formation of a diagnosis is selected, the following actions are taken: a criterion for the quality of the video image of the skin is set to ensure accurate analysis of the video image and diagnosis of melanoma; at least one recording of the dermatoscopic video image of the same skin region is repeated; the values of pixels of the recorded video image are measured; the recorded digital video images of the skin regions are superimposed; the values of pixels of the superimposed video image are measured; the average value of pixels of the superimposed video images is measured; the root mean square deviation of the values of pixels of the superimposed video images is measured, wherein said deviation characterises the noise level; the signal-to-noise ratio (S/N) is determined for the superimposed video images; the need to continue recording the video images of the skin is determined by comparing the resulting ratio (S/N) with the set criterion for the quality of the video image. Characteristic regions of the entire recorded image of the skin are then selected based on the attribute of the presence of intensity of anomalies; the threshold value of binarisation over the entire area of the image is determined; the adaptive threshold value of binarisation in the characteristic areas of the image is determined; binarisation is performed, and the end results are measured over the entire area of the recorded image by comparing the values of pixels to the threshold value. Binarisation is performed, and the end results are measured in the characteristic areas of the area of the image. The results of combined binarisation resulting from the superimposition by the "logic AND" operation of the measured results of image binarisation on the result of binarisation of the adaptive threshold are measured to reduce the background noise level and homogeneities within the area of the image of the melanoma. The image of combined binarisation image is subjected to dilatation. The measured combined binarised image is subjected to erosion to smooth the edges of the lines; the measured combined binarised image with the smoothed edges of the lines forming the skeleton of the lines is skeletonised; the skeletonisation of the measured combined binarised image is filtered, and the end results are measured to eliminate false branches of the skeleton lines.

EFFECT: increase in the objectivity, informative value, and accuracy in the course of diagnosis of skin melanoma.

5 cl, 5 dwg, 1 tbl