- Podcast
- 08/11/2025
- UK
Development And Evaluation Of Deep Learning Models For Detecting And Classifying Various Bone Tumours In Full-field Limb Radiographs Using Automated Object Detection Models
Description
In this episode of AI Talks with Bone & Joint, hosts Brian and Lisa delve into a groundbreaking study on the use of deep learning models for detecting and classifying bone tumors in full-field limb radiographs. The research, published in September 2025 by M Yamana and colleagues, investigates two major object detection models: DINO, an end-to-end model leveraging transformers, and YOLO, a well-known convolutional neural network model. Key highlights include the study's findings that DINO significantly outperforms YOLO in tumor detection rate (85.7% vs. 80.1%) and classification accuracy, especially for malignant tumors. The episode discusses the methodology behind the research, including the sample size of 642 tumors and the use of various performance metrics, and addresses limitations such as sample size and resolution constraints. Ultimately, the hosts emphasize the promising role of transformer-based models like DINO in enhancing diagnostic processes within healthcare, potentially increasing accuracy while easing the burden on medical professionals.
Part of: Surgical Techniques and Training Collection
"Development And Evaluation Of Deep Learning Models For Detecting And Classifying Various Bone Tumours In Full-field Limb Radiographs Using Automated Object Detection Models" is included in the following Surgical Techniques and Training playlist: