Conic cornea recognition method and system based on multi-dimensional feature adaptive fusion

Abstract

The invention discloses a keratoconus recognition method and system based on multi-dimensional feature self-adaptive fusion. The method comprises the following steps: (1) obtaining original corneal topographic map data of a plurality of independent samples by using a Pentacam anterior segment imaging system; (2) carrying out comprehensive judgment and labeling on the condition of the conic corneaon the five dimensions of each independent sample; (3) counting normal size, mean value, variance and extreme value information of original corneal topographic map data of five dimensions; (4) dividing a training set and a verification set; (5) processing topographic map data in the training set and the verification set; (6) constructing and training a residual convolutional neural network with five-dimensional feature adaptive fusion; (7) utilizing a Grad-CAM visualization mode to obtain average visualization information of the three types of test samples; and (8) performing prediction by using the trained model, and performing back propagation on the maximum prediction score to obtain a visual effect picture. According to the invention, the problem of poor recognition effect of the coniccornea in practical application can be solved.

Description

technical field [0001] The invention belongs to the technical field of medical data processing, and in particular relates to a keratoconus recognition method and system based on multi-dimensional feature adaptive fusion. Background technique [0002] In recent years, deep learning technology has developed rapidly in all walks of life, and medical assistance combined with deep learning technology has become the goal and direction of more and more people's efforts. Keratoconus is a slowly developing eye disease. It is difficult to diagnose in the early stage (that is, subclinical keratoconus), and there are no obvious clinical manifestations. The course of the disease varies from person to person. Significant corneal dilatation, central thinning, protruding forward, and conical shape are present, and advanced keratoconus can be diagnosed with a variety of ophthalmic equipment. The key problem is that in the early stage of the disease, it is difficult to diagnose but can be co...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above-mentioned problems existing in the prior art, the present invention provides a keratoconus recognition method and system based on multi-dimensional feature adaptive fusion, which can solve the problem of inaccurate prediction of keratoconus in practical applications, and can assist doctors in diagnosis

Images