A method and system for field-adaptive segmentation of oct retinal images

Abstract

The invention discloses a field-adaptive segmentation method and system for OCT retinal images. The method comprises the following steps: step 1, acquiring OCT retinal image data, and performing image preprocessing; step 2, combining feature extraction network, feature classification network and feature identification The network respectively constructs the intra-domain segmentation model and the cross-domain segmentation model; step 3, obtains the pseudo-label of the target domain according to the intra-domain segmentation model and the cross-domain segmentation model respectively; step 4, uses the self-selection active learning strategy to select effective pseudo-labels and original images, Add rich image data from the source domain, and train the cross-domain segmentation model again; step 5, use the trained cross-domain segmentation model to segment the OCT retinal image in the target domain. Through the technical solution of the invention, the generalization performance of the segmentation model can be effectively improved, reliable quantitative data can be provided for clinical ophthalmic disease diagnosis, and computer-aided personalized treatment can be realized.

Description

technical field [0001] The invention relates to the cross field of artificial intelligence and medical image processing, in particular to a field-adaptive segmentation method and system for OCT retinal images. Background technique [0002] Retinopathy is a common ophthalmic disease that seriously harms the human visual sensory system. Optical coherence tomography (OCT) imaging technology can clearly present each cell layer of the retina. Clinical ophthalmologists can diagnose diseases by observing the changes in the structure of retinal tissue layers, and the precise segmentation of retinal lesion areas in OCT images It is an important prerequisite and guarantee for disease diagnosis. Recently, deep convolutional neural networks have become a mainstream method for medical image segmentation and have been successfully applied to retinal image segmentation. However, due to the different imaging parameters of different manufacturers' equipment (such as Heidelberg, Zeiss, etc....

AI Technical Summary

Problems solved by technology

However, these methods all learn the common features of the source domain and the target domain, while ignoring the unique features of the target domain.

Images