SpineFM: Leveraging Foundation Models for Automatic Spine X-ray Segmentation

TL;DR

This paper presents SpineFM, a new pipeline that uses foundation models and the spine's geometry to automatically segment and identify vertebrae in spine X-ray images, achieving state-of-the-art accuracy.

Contribution

SpineFM introduces a novel inductive process leveraging foundation models and spine geometry for improved vertebral segmentation and identification in X-ray images.

Findings

Achieved 97.8% and 99.6% vertebrae identification on two datasets.

Segmentation Dice scores of 0.942 and 0.921, surpassing previous methods.

Demonstrated robustness of Medical-SAM-Adaptor for spine segmentation.

Abstract

This paper introduces SpineFM, a novel pipeline that achieves state-of-the-art performance in the automatic segmentation and identification of vertebral bodies in cervical and lumbar spine radiographs. SpineFM leverages the regular geometry of the spine, employing a novel inductive process to sequentially infer the location of each vertebra along the spinal column. Vertebrae are segmented using Medical-SAM-Adaptor, a robust foundation model that diverges from commonly used CNN-based models. We achieved outstanding results on two publicly available spine X-Ray datasets, with successful identification of 97.8\% and 99.6\% of annotated vertebrae, respectively. Of which, our segmentation reached an average Dice of 0.942 and 0.921, surpassing previous state-of-the-art methods.