Annotated Dataset for Knee Osteoarthritis Classification: 3-Class vs 5-Class KL Grading for Deep Learning Applications
收藏DataCite Commons2026-04-20 更新2026-05-04 收录
下载链接：
https://data.mendeley.com/datasets/cgjjbw8hsf/1
下载链接
链接失效反馈官方服务：
资源简介：
This dataset was curated to support the development and evaluation of deep learning models for automatic classification of knee osteoarthritis (KOA) severity from radiographic images. It enables comparative analysis between the conventional 5-class Kellgren–Lawrence (KL) grading system and a clinically consolidated 3-class scheme (Healthy, Mild–Moderate, Severe).
The dataset is derived from the “Knee Osteoarthritis Dataset with Severity Grading” (Kaggle) and has been reorganized to ensure methodological rigor. A patient-wise splitting strategy was implemented to prevent data leakage and ensure realistic generalization.
It contains 10,362 knee X-ray images distributed across training, validation, and independent test sets. Each image includes KL labels (0–4) and corresponding 3-class mappings.
Preprocessing includes resizing (300×300), normalization, and structured augmentation applied during training. Class imbalance was addressed using selective oversampling.
This dataset supports reproducible research in medical imaging and was used to develop an EfficientNetB3-based model optimized with Bayesian hyperparameter tuning and deployed in a web-based clinical decision support system (KneeAI).
提供机构：
Mendeley Data
创建时间：
2026-04-20
