Semi-supervised evaluation method and evaluation system based on dual consistency self-ensemble learning

Abstract

The invention provides a semi-supervised evaluation method and evaluation system based on dual consistency self-ensemble learning. The semi-supervised evaluation system based on dual consistency self-ensemble learning comprises a dual consistency average teacher framework module, an attention loss function module and an attention consistency loss function module. The invention designs a self-integration learning framework. The framework is composed of a student network and a teacher network which have the same structure. A new loss function based on an attention mechanism is designed to obtain an accurate attention area result; by performing dual consistency constraint on attention in lesion classification and positioning, the two networks can gradually optimize attention distribution and improve the performance of each other, training only depends on part of marked data and follows a semi-supervised training mode, a large amount of marked data is not needed, and the cost is low.

Description

technical field [0001] The invention relates to the field of disease diagnosis based on artificial intelligence methods, in particular to a semi-supervised evaluation method and evaluation system based on double consistency self-integrated learning. Background technique [0002] Osteoarthritis (OA) of the knee is one of the most common joint diseases characterized by lack of integrity of articular cartilage and concurrent changes in subchondral bone and joint structure. Knee osteoarthritis, if not treated early, can lead to joint necrosis and even disability. Knee articular cartilage defect, as a prominent feature of knee osteoarthritis, is highly related to the onset of arthritis. Therefore, it is necessary to quantitatively evaluate knee articular cartilage defect in the early stage. Today, deep convolutional neural networks (CNNs) have achieved great success in the field of computer-aided diagnosis. However, the construction of deep learning models usually requires a la...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a semi-supervised evaluation method and evaluation system based on double consistency self-integrated learning, to solve the usual needs of the construction of deep learning models for quantitative evaluation of knee osteoarthritis in existing artificial intelligence methods A large amount of labeled data is a technical problem with high cost

Images