A Cataract Detection Method Based on Improved YOLO11n
DOI: https://doi.org/10.62517/jbdc.202601120
Author(s)
Yixiang Shang1, Luping Qian1, Haoran Li1, Mengqi Liu1, Mingyue Xue2, Mingxu Li3
Affiliation(s)
1School of Artificial Intelligence and Big Data, Henan University of Technology, Zhengzhou, Henan, China
2School of Economics and Trade, Henan University of Technology, Zhengzhou, Henan, China
3IFLYTEK Co., Ltd., Hefei, Anhui, China
Abstract
Cataract is renowned as the world's leading cause of blindness. Current YOLO series models suffer from deficiencies such as severe feature interference, low detection rate for small lesions, and significant impact of artifacts. To address these issues, this paper proposes the ECL-YOLO11 model based on an improved YOLO11 approach. YOLO11 is divided into two layers: feature extraction and feature fusion. In the feature extraction stage, the backbone network is replaced with EfficientViT to enhance global feature modeling capability, and C3k2 is substituted with the C3k2_ContextGuided module to strengthen multi-scale context acquisition. In the feature fusion stage, SPPF is replaced with SPPF_LSKA to optimize micro-lesion response. In cataract detection experiments, the ECL-YOLO11n model achieves an accuracy of 95.52%, recall of 99.79%, F1-Score of 97.60%, mAP0.5 of 99.35%, and mAP0.5-0.95 of 77.46%. Compared with the YOLO11n baseline model, these metrics are improved by 2.34 to 3.7 percentage points. Compared with mainstream models such as Faster R-CNN and YOLOv8n, it shows varying degrees of improvement in classification metrics and an increase of 1.15 to 1.16 percentage points in target detection metrics. The mAP0.5 values for "no cataract" and "with cataract" samples are 99.46% and 99.20% respectively.
Keywords
Cataract Screening; EfficientViT; C3k2_ContextGuided; SPPF_LSKA; YOLO11
References
[1]Yang Bin, Chen Yi, Zong Yixin. Modern Light Environment, Visual Health Problems and Research Status. Science and Technology China, 2020, (02):15-17.
[2]Ren Shaoqing, He Kaiming, Girshick Ross & Sun Jian. (2017). Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE transactions on pattern analysis and machine intelligence, 39(6), 1137-1149.https://doi.org/10.1109/ TPAMI.2016.2577031.
[3]Li Zhenwei. Improvement and System Design of Strawberry Fruit Detection Method Based on Deep Learning. Nanjing University of Information Science and Technology, 2025. DOI:10.27248/d.cnki. gnjqc.2025.000895.
[4]Sun Bin. Research on Cataract Auxiliary Detection and Surgical Image Segmentation Algorithm Based on Deep Learning. Taiyuan University of Technology, 2024. DOI:10.27352/d.cnki.gylgu.2024.001586.
[5]Zhang Linlin. Research on Automatic Cataract Classification Based on Deep Learning. Beijing University of Technology, 2018.
[6]Wang Yong. Ultrasound Cataract Detection Algorithm Based on Deep Learning. Modern Computer, 2021, (11):97-101.
[7]Chen Sihan, Liu Yong, He Xiang. Factory Pedestrian Detection Algorithm Based on Improved YOLOv8. Modern Electronics Technique, 2024, 47(24):160-166. DOI: 10.16652/j.issn.1004-373x.2024.24.025.
[8]Zhang Xinran, Wang Xinzhong, Zhou Kuizhou, et al. Multi-Target Dairy Cow Daily Behavior Detection in Complex Environments Based on Improved YOLO11n. Transactions of the Chinese Society of Agricultural Engineering, 2025, 41(14):155-164.
[9]Xiang Houxue, Xu Guiyang, Zhang Yuhua, et al. Intelligent Recognition Algorithm for Typical Rail Defects Based on Improved YOLOv8. Science Technology and Engineering, 2025, 25(18):7785-7792.
[10]Tan M, Le Q. Efficientnet: rethinking model scaling for convolutional neural networks // International Conference on Machine Learning. Chambridge: PMLR, 2019: 6105-6114.
[11]Wu Shichen, Mao Yuming, Hu Huizhong, et al. Grape Leaf Disease Detection Method Based on Improved YOLO11n. Transactions of the Chinese Society of Agricultural Engineering, 2025, 41(14):140-147.
[12]Zhang Z, PANG Y. CGNet: cross-guidance network for semantic segmentation. Science China (Information Sciences), 2020, 63(02):49-64.
[13]Li Jun, Zhou Keyu, Zou Jun, et al. Detection Algorithm for Protective Equipment Wearing in Construction Scenarios Based on Improved YOLOv8n. Journal of Zhengzhou University (Engineering Science Edition), 2025, 46(03):19-25+104. DOI:10.13705/j. issn.1671- 6833.2025.03.002.
