Multi‐label Ensemble Model for Knee Joint Anatomy and Lesion Segmentation: Segmentation of Clinical Images With Ensembling, Morphology, and Attention (SCEMA)

Key Points

• This research aims to develop a fully automatic model for segmenting knee joint anatomy and lesions in MRI images.
• Utilized a multi-label ensemble model for segmentation tasks.
• Applied morphological techniques to improve segmentation quality.
• Incorporated attention mechanisms to enhance focus on relevant anatomical structures.
• Achieved high-quality segmentation of knee joint anatomy and lesions.
• Segmentation results were validated and reproducible across different cases.
• BMEL quality labels and segmentation annotations were made publicly available for future research.