Hippocampal Functional Radiomic Features for Identification of the Cognitively Impaired Patients from Low-Back-Related Pain: A Prospective Machine Learning Study

Abstract

Purpose: To investigate whether functional radiomic features in bilateral hippocampi can identify the cognitively impaired patients from low-back-related leg pain (LBLP). Patients and Methods: For this retrospective study, a total of 95 clinically definite LBLP patients (40 cognitively impaired patients and 45 cognitively preserved patients) were included, and all patients underwent functional MRI and clinical assessments. After calculating the amplitude of low-frequency fluctuations (ALFF), regional homogeneity (ReHo), voxel-mirrored homotopic connectivity (VMHC) and degree centrality (DC) imaging, the radiomic features (n = 819) of bilateral hippocampi were extracted from these images, respectively. After feature selection, machine learning models were trained. Finally, we further analyzed the relationship between the hippocampal functional radiomic features and clinical measures, to explore the clinical significance of these features. Results: The combined radiomic features model logistic regression algorithm superior performance in distinguishing cognitively impaired patients from LBLP (AUC = 0.970, accuracy = 92.3%, sensitivity = 92.3%, specificity = 92.3%) compared to the other models. Additionally, radiomic wavelet features were correlated with Montreal Cognitive Assessment (MoCA) and Hamilton Anxiety Scale, present pain intensity scores in cognitively impaired LBLP patients (P < 0.05, with Bonferroni correction). Conclusion: Hippocampal functional radiomic features are valuable for diagnosing cognitively impaired patients from LBLP.