Background: Spinopelvic parameters (SPP), derived from vertebral body measurements, are crucial in informing pre-operative spine surgery planning. Current methods of calculating SPP are labor-intensive and imprecise with clinicians using archaic software to manually draw lines on spine x-rays. By leveraging artificial intelligence and computer vision, SPP can be calculated more efficiently and consistently with a standardized approach. This would reduce physician burden and improve patient outcomes. 

Methods: 167 sagittal spine x-rays were collected and randomly split into a training set (67 images) and a testing set (100 images). A total of 536 images were produced for the training set through image augmentation on the original 67 x-rays. All images were manually annotated to create binary segmentation masks. A U-Net model, a convolutional neural network widely used for bio-medical images, will be trained on this dataset with a goal to automate the segmentation of 25 vertebral bodies (C2-L5, sacrum, femoral head). Once all bodies are segmented, computer vision edge-detection algorithms have been implemented to automatically calculate SPP. Finally, an intuitive web application for spinal surgeons was developed to help visualize SPP on the spine with overlayed illustrations. 

Expected Results: The goal is to create a web application that allows a surgeon to upload a sagittal spine x-ray of their patient and receive immediate SPP without extra tedious labor. This process would provide all SPP in seconds and remove any possibility for human error. With computer vision and web interface aspects already implemented, further success of the application relies on performance of the model still in training.  

Further Impact: A fully functioning application would assist in pre-operative planning, post-operative assessment, and avoid high potential for error which can have a large impact on surgical success. It would also accelerate spinal deformity research and advance neurosurgical education with informative visualizations.