METHOD FOR THREE-DIMENSIONAL VISUALIZATION, QUANTITATIVE ASSESSMENT AND MONITORING OF DYNAMICS OF CHANGES IN VOLUMES OF PATHOLOGICAL STRUCTURES ON OPTICAL COHERENCE TOMOGRAPHY SCANS OF HUMAN RETINA USING ARTIFICIAL INTELLIGENCE ALGORITHM Russian patent published in 2025 - IPC A61B8/10 A61B3/10 G06T7/10 G06N3/02 

FIELD: medicine; physics.

SUBSTANCE: invention relates to medical decision support systems. Disclosed is a method in which an optical coherence tomograph is used to scan the retina. As a result, a series of scans of structural OCT is obtained, which makes a macular cube in DICOM format, which also contains information on the interval between scans in micrometres and information on the scanning area in millimetres. Then, using deep convolutional neural networks on each OCT scan of the retina from this set, at least 9 different classes of pathological structures are detected: intraretinal cysts, subretinal fluid, retinal pigment epithelium detachment, subretinal hyperreflective material, retinal drusen, epiretinal membrane, vitreomacular traction (includes two subclasses: posterior hyaloid membrane and vitreomacular attachment area), through macular rupture, lamellar macular rupture. Obtained quantitative and qualitative data are used to form a differential diagnostic search algorithm. Differential diagnostic search algorithm consists in logical search of the most probable pathology or their combination based on the presence and combination of pathological structures on the OCT scans. Obtained segmented masks are used to form a three-dimensional visualization of the detected pathological structures while maintaining a relative spatial orientation by means of an algorithm for constructing surfaces of the pathological structures. Three-dimensional visualization of pathological structures enables to control the display settings of each type of pathological structures, which is also an important functional of the tool for displaying spatial characteristics of pathological structures. Calculation of quantitative estimation of volumes of pathological structures is realized by means of their quantitative estimation algorithm. Further, an algorithm for quantitative assessment of volumes of the following pathological structures is implemented: intraretinal cysts, subretinal fluid, retinal pigment epithelium detachment, retinal drusen, subretinal hyperreflective material. Based on the calculated quantitative assessment of the volumes of pathological structures and information on the volumes of these structures based on previous scans of the retina, an algorithm for monitoring the dynamics of changes in the volume of the pathological structures is implemented.

EFFECT: invention provides auxiliary information, in particular detection of signs of diabetic retinopathy in patients with diabetes mellitus, determination of stage of disease and evaluation of effectiveness of treatment, as well as determination of risk of diabetic macular oedema.

7 cl, 14 dwg, 1 tbl