Exploring the Relationship between Severity of Intervertebral Disc Degeneration and Dyslipidemia: A Deep Learning Approach

Background: Intervertebral disc degeneration (IDD) is a chronic, noninflammatory, degenerative process that occurs in the lumbar spine and surrounding soft tissues. Although dyslipidemia has been proven to be related to degenerative changes in the lumbar spine, its relationship with IDD severity of IDD is not been fully explored and elucidated. This study employed deep learning to improve IDD prediction accuracy and suggest new preventive and therapeutic strategies for clinical application. Methods: We enrolled 369 IDD patients. Serum was collected and analyzed for high-density lipoprotein (HDL-C), low-density lipoprotein (LDL-C), lipoprotein (a), total cholesterol (TC), triglycerides (TG), apolipoprotein A1, apolipoprotein B (APO-B), apolipoprotein E (ApoE), lipoprotein (a) (LP(a)) levels, Apo B/Apo A1 ratio, and other baseline data. IDD was assessed according to the Pfirrmann grading system, and Grades I-III were categorized as low-grade disc degeneration, while Grades III-IV were considered moderate grade, Grades IV-V were considered high grade. We used a deep learning model with ProbSparse Self-Attention and Kolmogorov-Arnold Networks (KAN) to predict the degeneration scores. Moreover, we calculated the SHapley Additive ExPlanations (SHAP) values of the features to evaluate their importance. Results A total of 50 patients are enrolled in the low degeneration group, 179 in the moderate degeneration group, and 165 in the high degeneration group. According to the univariate analysis of the different degeneration groups, there are statistically significant differences in terms of age, low-density lipoprotein, apolipoprotein B, and Apo B/Apo A1 ratio (P <0.001, P = 0.036, and P = 0.046, P = 0.032, respectively, which were smaller than 0.05), but no statistically significant differences in terms of high-density lipoprotein, lipoprotein (a), total cholesterol (TC), triglycerides (TG), apolipoprotein A1, gender, fasting blood glucose concentration, and length of hospital stay (P =0.86, P =0.532, P =0.359, P =0.188, P =0.702, and P =0.348). By employing a deep learning model to assess the influence of various factors on disease severity, and examining the impact of four statistically significant factors (age, LDL, Apo B, and Apo B/Apo A1 ratio) on the severity of lumbar disc degeneration among the three groups, it is determined that age and cholesterol serve as predictors for the severity of intervertebral disc degeneration (IDD).Grouping patients by disc degeneration severity reveals that age significantly influences the moderate degeneration group, whereas Apo B and the Apo B/Apo A1 ratio are most impactful in the high degeneration group.compared between the moderate and high degeneration groups, when LDL-C values are the same, its impact is the greatest in the moderate degeneration group. Conclusion: Age significantly correlates with intervertebral disc degeneration, while the ratios of low-density lipoprotein, apolipoprotein B, and Apo B/Apo A1 are associated with varying degrees of disc degeneration across different groups.Age and cholesterol levels serve as indicators for assessing the severity of intervertebral disc degeneration.