Line-based Instance Segmentation for Robust and Automatic MRI Spine Scan Planning

Motivation: To enhance the accuracy and efficiency of automated spine view planning, this approach can be seamlessly adapted for MRI instruments operating at various field strengths. Goal(s): To develop an intelligent prescription method for transverse scanning planes of intervertebral discs (IVDs) on sagittal spine images, applicable across various magnetic field strengths. Approach: We employed an instance segmentation approach for the detection and segmentation of intervertebral discs (IVDs). The network is fine-tuned utilizing a weakly supervised method on images without masks. Results: We achieve improved plane prescription on images acquired from MRI instruments of different field strength and at different spine location. Impact: Our method improved the accuracy and speed of spine positioning, providing more precise spinal transection scanning results. This can help doctors in diagnosing lumbar spine diseases more effectively and improve the diagnosis experience of patients.