Google-MedGemma Based Abnormality Detection in Musculoskeletal radiographs
This work addresses clinical radiograph triage by providing scalable and accurate abnormality detection, though it is incremental as it applies an existing foundation model to a specific medical imaging task.
The paper tackled automatic abnormality detection in musculoskeletal radiographs using a MedGemma-based framework, achieving strong performance that exceeded conventional convolutional and autoencoder-based metrics.
This paper proposes a MedGemma-based framework for automatic abnormality detection in musculoskeletal radiographs. Departing from conventional autoencoder and neural network pipelines, the proposed method leverages the MedGemma foundation model, incorporating a SigLIP-derived vision encoder pretrained on diverse medical imaging modalities. Preprocessed X-ray images are encoded into high-dimensional embeddings using the MedGemma vision backbone, which are subsequently passed through a lightweight multilayer perceptron for binary classification. Experimental assessment reveals that the MedGemma-driven classifier exhibits strong performance, exceeding conventional convolutional and autoencoder-based metrics. Additionally, the model leverages MedGemma's transfer learning capabilities, enhancing generalization and optimizing feature engineering. The integration of a modern medical foundation model not only enhances representation learning but also facilitates modular training strategies such as selective encoder block unfreezing for efficient domain adaptation. The findings suggest that MedGemma-powered classification systems can advance clinical radiograph triage by providing scalable and accurate abnormality detection, with potential for broader applications in automated medical image analysis. Keywords: Google MedGemma, MURA, Medical Image, Classification.