An auto-Segmentation pipeline for Diffusion Tensor Imaging on spinal cord

Diffusion Tensor Imaging (DTI) is a crucial imaging technique for accurately assessing pathological changes in cervical spondylotic myelopathy (CSM). However, the segmentation of spinal cord DTI images primarily relies on manual methods, which are labor-intensive and heavily dependent on the subjective experience of clinicians. Which has posed significant challenges for clinicians who use DTI based analysis in clinical decision making. This study developed a clinical application orientated auto segmentation pipeline to deliver AI-driven segmentation measurements for spinal cord DTI. The study involved a dataset of 89 CSM patients, with each patient’s DTI comprising 12 slices, totaling 1,059 DTI slices. A novel deep learning-based model, named as SCS-net (spinal cord segmentation net), was designed with U-shape encoder-decoder and a focal loss function to address class imbalance. The performance of the SCS-net was evaluated using both general segmentation metrics (precision, recall, and Dice coefficient) and DTI-specific metrics based on FA values. The automatic precise segmentation tool divided the spinal cord DTI into 8 regions, i.e. the lateral, dorsal, ventral, and gray matter areas on the left and right sides The model demonstrated satisfactory accuracy, with error rates in FA values were evaluated as 5.32%, 10.14%, 7.37%, and 5.70% on the left side, and 4.60%, 9.60%, 8.74%, and 6.27% on the right side of the spinal cord, respectively, affirming the model’s consistent performance in radiological rationality. The pipeline was designed to be user-friendly, allowing clinicians to easily access automated segmentation outputs through simple command operations. In conclusion, the proposed AI-driven segmentation pipeline significantly reduces the dependence on manual interpretation, providing a scalable solution that can streamline clinical workflows and improve potential diagnostic outcomes for patients.