Recognition methods based on artificial neural networks

By optimizing the multi-layer artificial neural network module and the total loss function, the accuracy problem of convolutional neural networks in identifying fundus lesions was solved, and efficient identification of lesion areas was achieved.

CN115511861BActive Publication Date: 2026-07-17SHENZHEN SIBRIGHT TECH CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHENZHEN SIBRIGHT TECH CO LTD
Filing Date
2020-11-27
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing convolutional neural networks tend to overlook lesion areas with low attention when identifying lesions such as those in the fundus, resulting in low accuracy in tissue lesion identification.

Method used

A multi-layer artificial neural network module is adopted, including a first artificial neural network for feature extraction, a second artificial neural network for generating attention heatmaps and complementary attention heatmaps, and a third artificial neural network for recognition. The network is optimized by a total loss function, and the recognition accuracy of lesion areas is improved by combining attention mechanism and complementary attention mechanism.

Benefits of technology

It improves the accuracy of tissue lesion identification, especially in fundus lesions, enhancing the identification effect of lesion areas and reducing misjudgment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115511861B_ABST
    Figure CN115511861B_ABST
Patent Text Reader

Abstract

This disclosure describes a recognition method based on an artificial neural network, comprising: acquiring a tissue image; receiving the tissue image and performing lesion recognition and training on the tissue image using an artificial neural network module; recognizing the examination image based on a feature map and an attention heatmap to obtain a recognition result; combining the recognition result with a labeled image to obtain a first loss function without using an attention mechanism; combining the recognition result with the labeled image to obtain a second loss function with an attention mechanism; using the first loss function and the second loss function to obtain a total loss function including a first loss term based on the first loss function and a second loss term based on the difference between the second loss function and the first loss function; and using the total loss function to optimize the artificial neural network module, thereby effectively improving the recognition rate of tissue lesions.
Need to check novelty before this filing date? Find Prior Art