368 Preoperative Soft Tissue Imaging Features Elucidated by Artificial Intelligence May Best Predict Radiographic Proximal Junction Kyphosis Following Adult Spinal Deformity Surgery

INTRODUCTION: Proximal junction kyphosis (PJK) following adult spinal deformity surgery (ASD) remains difficult to predict. In a cohort of patients undergoing ASD surgery, we used artificial intelligence to compare three models of preoperatively predicting PJK using: 1) traditional radiographic measurements and demographics 2) raw preoperative scoliosis x-rays, and 3) raw preoperative thoracic T1 MRIs. METHODS: A single-institution retrospective cohort study was undertaken for patients undergoing ASD surgery from 2009-21. PJK was defined as a sagittal Cobb angle of upper-instrumented vertebra (UIV) and UIV+2>10° and a postoperative change in UIV/UIV+2>10°. For Model-1, A support vector machine was used to predict PJK within 2 years postoperatively using clinical and traditional sagittal/coronal radiographic variables and intended levels of instrumentation. Next, for Model-2, a convolutional neural network (CNN) was trained on raw preoperative lateral/posterior-anterior scoliosis x-rays. Finally, for Model-3, a CNN was trained on raw thoracic T1 MRIs. RESULTS: A total of 187 patients underwent ASD surgery with at least 2-year follow-up and 89 (47.6%) developed radiographic PJK within 2 years. Model-1: Using clinical variables and traditional radiographic measurements, the support vector machine predicted PJK with a sensitivity:57.2% and specificity:56.3%. Model-2: a CNN with raw scoliosis x-rays predicted PJK with sensitivity:68.2% and specificity:58.3%. Model-3: a CNN with raw thoracic MRIs predicted PJK with sensitivity:73.1% and specificity:79.5%. Finally, an attention map outlined the imaging features used by Model-3 to predict PJK. Soft tissue features predominated all true positive PJK predictions. CONCLUSIONS: The use of raw MRIs in an artificial intelligence model improved the accuracy of PJK prediction compared to raw scoliosis X-rays and traditional clinical/radiographic measurements. The improved predictive ability using MRI may indicate that PJK is best predicted by soft-tissue degeneration and muscle atrophy.