Data-Driven Framework for Classification and Surgical Planning of Spinal Deformity

点击翻译按钮获取中文摘要

PROJECT SUMMARY Adolescent idiopathic scoliosis (AIS) impacts 2-4% of the adolescent population. AIS causes a three- dimensional deformity of the spinal column affecting the patients’ normal motion and posture and may cause lung and heart dysfunction, early onset osteoarthritis, and disc degeneration if left untreated. Spinal fusion surgery in progressive cases of scoliosis remains the main treatment option. The variation in patients’ pre-operative characteristics, the surgical implants, and the surgical maneuvers have resulted in a wide range of surgical outcomes, 20% of which remains to be less than satisfactory. As the suboptimal surgical outcomes can significantly impact the cost, risk of revision surgery, and long-term rehabilitation of the adolescent patients, objective patient-specific models that can predict the outcome of different surgical treatment scenarios and determine the optimal surgical intervention for individuals are of critical need. The central hypothesis of the proposed work is that identifying the key features of a 3D spinal curve before the operation and the intraoperative surgical interventions the influence the long-term outcomes can provide a quantitative framework for predicting the surgical outcomes in this patient population. To this end, we propose (i) to identify the patient-specific and surgeon modifiable predictors of the spinal fusion outcomes in an in-house database of surgical AIS patients using machine learning, (ii) to develop a probabilistic predictive model of the outcomes as a function of pre-operative patient condition and the surgical interventions and (iii) to develop a fully automated framework that allows online image processing and assigns a treatment option that probabilistically determines the surgical outcome for a new patient based on a prior learning algorithm. The innovation of this approach is in developing the first data-driven predictive model for surgical planning of AIS patients that allows comparing different treatment scenarios through a probabilistic predictive framework and recommending surgical intervention that leads to an optimal outcome for a given patient. This knowledge-based algorithm automatically extracts the spinal curve patterns from the medical images as a classifier. The exploitation of an automated image processing algorithm to develop a reduced ordered model of the spinal deformity allows a fast quantitative analysis appropriate for direct clinical dissemination. It is aimed to use this model as an assistive tool for personalized surgical decision making of the AIS patients in the clinical setups. This assistive tool, which will be trained and tested using a large database of the medical images of the AIS patients, can make significant contribution to the field by developing a quantitative approach that considers a combinations of surgical methods and provides recommendations to achieve an improved outcome of the spinal deformity surgery in the pediatric population.