Evaluation of the Predictors for Unfavorable Clinical Outcomes of Degenerative Lumbar Spondylolisthesis After Lumbar Interbody Fusion Using Machine Learning

Background: An increasing number of geriatric patients are suffering from degenerative lumbar spondylolisthesis (DLS) and need a lumbar interbody fusion (LIF) operation to alleviate the symptoms. Our study was performed aiming to determine the predictors that contributed to unfavorable clinical efficacy among patients with DLS after LIF according to the support vector machine (SVM) algorithm. Methods: A total of 157 patients with single-segment DLS were recruited and performed LIF in our hospital from January 1, 2015 to October 1, 2020. Postoperative functional evaluation, including ODI and VAS were, performed, and endpoint events were defined as significant relief of symptom in the short term (2 weeks postoperatively) and long term (1 year postoperatively). General patient information and radiological data were selected and analyzed for statistical relationships with the endpoint events. The SVM method was used to establish the predictive model. Results: Among the 157 consecutive patients, a postoperative unfavorable clinical outcome was reported in 26 patients (16.6%) for a short-term cohort and nine patients (5.7%) for a long-term cohort. Based on univariate and multivariate regression analysis, increased disc height (DH), enlarged facet angle (FA), and raised lateral listhesis (LLS) grade were confirmed as the risk factors that hindered patients' short-term functional recovery. Furthermore, long-term functional recovery was significantly associated with DH alone. In combination with the SVM method, a prediction model with consistent and superior predictive performance was achieved with average and maximum areas under the receiver operating characteristic curve (AUC) of 0.88 and 0.96 in the short-term cohort, and 0.78 and 0.82 in the long-term cohort. The classification results of the discriminant analysis were demonstrated by the confusion matrix. Conclusions: The proposed SVM model indicated that DH, FA, and LLS were statistically associated with a clinical outcome of DLS. These results may provide optimized clinical strategy for treatment of DLS

Cohort · Lumbar · Machine learning · Multivariate analysis · Multivariate statistics · Receiver operating characteristic · Spondylolisthesis · Univariate · Univariate analysis · Cervical and Thoracic Myelopathy · Medical Imaging and Analysis · Medicine · Spine and Intervertebral Disc Pathology · Internal Medicine · Surgery

• Deep learning and medical image processing for coronavirus (Covid-19) pandemic

  Open Access•Sweta Bhattacharya, Praveen Kumar Reddy Maddikunta et al.•Sustainable Cities and Society•2021

  Add to list