Abnormal Spontaneous Brain Activity in Occupational Chronic Benzene Poisoning: A Resting-State fMRI Study with Machine Learning.
Authors
Affiliations (7)
Affiliations (7)
- School of Medical Imaging, Shandong Medical And Pharmaceutical University, Yantai,264000,China. Electronic address: [email protected].
- Yantaishan Hospital Affiliated to Binzhou Medical University, PR China. Electronic address: [email protected].
- Yantaishan Hospital Affiliated to Binzhou Medical University, PR China. Electronic address: [email protected].
- School of Medical Imaging, Shandong Medical And Pharmaceutical University, Yantai,264000,China. Electronic address: [email protected].
- School of Medical Imaging, Shandong Medical And Pharmaceutical University, Yantai,264000,China. Electronic address: [email protected].
- School of Medical Imaging, Shandong Medical And Pharmaceutical University, Yantai,264000,China. Electronic address: [email protected].
- Yantaishan Hospital Affiliated to Binzhou Medical University, PR China. Electronic address: [email protected].
Abstract
This study aimed to investigate abnormal spontaneous brain activity in patients with occupational chronic benzene poisoning (OCBP) using resting-state functional magnetic resonance imaging (rs-fMRI) and to explore the within-sample discriminative value of rs-fMRI features. 20 patients with OCBP and 20 age- and sex-matched healthy controls (HCs) underwent rs-fMRI scanning and neuropsychological assessments. Amplitude of low- frequency fluctuations (ALFF), fractional amplitude of low-frequency fluctuations (fALFF), and regional homogeneity (ReHo) were used to assess local spontaneous brain activity. Group comparisons and correlation analyses were performed. Imaging features from significantly altered regions were selected using Least Absolute Shrinkage and Selection Operator (LASSO) and used to construct Support Vector Machine (SVM) classification models. Model performance was evaluated by leave- one-out cross-validation. Compared with HCs, OCBP patients showed widespread abnormalities in brain regions associated with the default mode network, salience network, and executive control network. The right precuneus exhibited alterations across all three indices. Six fALFF-symptom associations had unadjusted P values below 0.05, but none remained significant after correction across 70 tests (minimum q = 0.196). The combined model yielded an accuracy of 77.5% and an AUC of 0.873 (95% CI, 0.742-0.966), numerically exceeding the single-measure models; right-precuneus ALFF and fALFF features had the largest weights. OCBP patients display extensive spontaneous brain activity abnormalities involving multiple networks and the limbic system, with the right precuneus as a key region. Multi-indicator rs-fMRI-based machine learning models show good classification performance, but the symptom associations and classifier require independent validation in larger, exposure-characterized cohorts.