Automatic Lung Segmentation in Chest X-Ray Images Using SAM With Prompts From YOLO

Despite the impressive performance of current deep learning models in medical imaging, transferring the lung segmentation task in X-ray images to clinical practice is still a pending task. This study evaluated the performance of a fully automatic framework for lung field segmentation in chest X-ray images. The framework is rooted in combining the Segment Anything Model (SAM) with prompt capabilities, and the You Only Look Once (YOLO) model to provide effective prompts. Transfer learning, loss functions, and several validation strategies were thoroughly assessed. This provided a complete benchmark that enabled future research studies to fairly compare new segmentation strategies. The results achieved demonstrated significant robustness and generalization capability against the variability in sensors, populations, disease manifestations, device processing, and imaging conditions. The proposed framework was computationally efficient, could address bias in training over multiple datasets, and had the potential to be applied across other domains and modalities.

This work was supported by Consejería de Universidad, Investigación e Innovación de la Junta de Andalucía, under Grant ProyExcel_00942.