I-Corps: A machine learning algorithm to predict recurrent disc herniation following microdiscectomy surgery

点击翻译按钮获取中文摘要

The broader impact/commercial potential of this I-Corps project is the development of software to predict spine surgery complications. There are 1.2 million spine surgeries nationally per year and about 25% (300,000 cases) of these are microdiscectomies. The main goal of microdiscectomy is to take pressure off your nerves to relieve back pain. Among spinal surgeries, microdiscectomies are less invasive and have quicker recovery times, however, 5-10% of microdiscectomy patients reherniate leading to additional surgery costs, longer patient recovery times, and lost productivity. Additionally, over 1/3 of reherniations occur within 3 months of the initial operation. To put this in perspective, these patients spend 6 weeks recovering from their first operation only to reherniate within a few months and return to the operating table for a second microdiscectomy or a spinal fusion. These revision surgeries often require months of further recovery time and pain. These secondary surgeries are also challenging for hospitals and insurance companies. Re-admissions decrease hospital quality ratings, and are often not reimbursed by insurance companies if they happen within 90 days. The proposed technology helps support a reduction in surgical re-admittance by identifying patients likely to reherniate and allowing for risk mediation, saving patients unnecessary pain and reducing hospital costs. This I-Corps project is based on the development of a machine learning algorithm that uses presurgical data to predict which patients are likely to suffer from recurrent disc herniation following microdiscectomy surgery. This technology aims to reduce the complication rates, thereby increasing patient satisfaction and decreasing costs to the patient and hospital. The machine learning algorithm is capable of correctly identifying 98% of herniation patients as either at risk or not at risk of reherniation in a cohort of 350 patients from one institute. Expansion of this work to multiple institutes is underway. Results have been collected from 1077 patients from 4 institutes in 3 countries thus far, showing that predicted rates of reherniation are higher in reherniated patients; however, the overall percent correct classification is still poor in some institutes because of physician inter-evaluator variability in calculating input metrics from radiographs. The goal is to develop semi-automated image analysis software to calculate input metrics, creating more consistency among institutions. This consistency will make it possible to provide a predictive tool capable of compiling all of the potential risk factors for reherniation and to report a single unified probability of risk so that care decisions are better informed.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.