A Cross-Domain Weakly Supervised Diabetic Retinopathy Lesion Identification Method Based on Multiple Instance Learning and Domain Adaptation
Open Access
- 20 September 2023
- journal article
- research article
- Published by MDPI AG in Bioengineering
- Vol. 10 (9) , 1100
Abstract
Accurate identification of lesions and their use across different medical institutions are the foundation and key to the clinical application of automatic diabetic retinopathy (DR) detection. Existing detection or segmentation methods can achieve acceptable results in DR lesion identification, but they strongly rely on a large number of fine-grained annotations that are not easily accessible and suffer severe performance degradation in the cross-domain application. In this paper, we propose a cross-domain weakly supervised DR lesion identification method using only easily accessible coarse-grained lesion attribute labels. We first propose the novel lesion-patch multiple instance learning method (LpMIL), which leverages the lesion attribute label for patch-level supervision to complete weakly supervised lesion identification. Then, we design a semantic constraint adaptation method (LpSCA) that improves the lesion identification performance of our model in different domains with semantic constraint loss. Finally, we perform secondary annotation on the open-source dataset EyePACS, to obtain the largest fine-grained annotated dataset EyePACS-pixel, and validate the performance of our model on it. Extensive experimental results on the public dataset FGADR and our EyePACS-pixel demonstrate that compared with the existing detection and segmentation methods, the proposed method can identify lesions accurately and comprehensively, and obtain competitive results using only coarse-grained annotations.Keywords
Funding Information
- Technological Innovation 2030—“New Generation Artificial Intelligence” Major Project (2022ZD0115902, 2022ZD0161902, 2019-I2M-5-016)
- CAMS Innovation Fund for Medical Sciences (2022ZD0115902, 2022ZD0161902, 2019-I2M-5-016)
This publication has 24 references indexed in Scilit:
- A Comparison of Five Multiple Instance Learning Pooling Functions for Sound Event Detection with Weak Labeling2019
- Multiple instance learning for histopathological breast cancer image classificationExpert Systems with Applications, 2019
- A Framework for Identifying Diabetic Retinopathy Based on Anti-noise Detection and Attention-Based Fusion2018
- Deep Learning on Low-Resource DatasetsPreprints.org, 2018
- Indian Diabetic Retinopathy Image Dataset (IDRiD): A Database for Diabetic Retinopathy Screening ResearchData, 2018
- Lesion Detection and Grading of Diabetic Retinopathy via Two-Stages Deep Convolutional Neural Networks2017
- Zoom-in-Net: Deep Mining Lesions for Diabetic Retinopathy Detection2017
- Deep Residual Learning for Image Recognition2016
- Learning Deep Features for Discriminative Localization2016
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015