Early-stage diagnosis of HIV-associated neurocognitive disorders via multiple learning models based on resting-state functional magnetic resonance imaging.
Authors
Affiliations (4)
Affiliations (4)
- Department of Radiology, Beijing Youan Hospital, Capital Medical University, Beijing, China.
- Department of Neuro-oncology Cancer Center, Beijing Tiantan Hospital, Capital Medical University, Beijing, China.
- Beijing Advanced Innovation Centre for Biomedical Engineering, Beihang University, Beijing, China.
- Laboratory for Clinical Medicine, Capital Medical University, Beijing, China.
Abstract
People living with human immunodeficiency virus (PLWH) are at risk of human immunodeficiency virus (HIV)-associated neurocognitive disorders (HAND). The mildest disease stage of HAND is asymptomatic neurocognitive impairment (ANI), and the accurate diagnosis of this stage can facilitate timely clinical interventions. The aim of this study was to mine features related to the diagnosis of ANI based on resting-state functional magnetic resonance imaging (rs-fMRI) and to establish classification models. A total of 74 patients with 74 ANI and 78 with PLWH but no neurocognitive disorders (PWND) were enrolled. Basic clinical, T1-weighted imaging, and rs-fMRI data were obtained. The rs-fMRI signal values and radiomics features of 116 brain regions designated by the Anatomical Automatic Labeling template were collected, and the features were selected via the least absolute shrinkage and selection operator. rs-fMRI, radiomics, and combined models were constructed with five machine learning classifiers, respectively. Model performance was evaluated via the mean area under the curve (AUC), accuracy, sensitivity, and specificity. Twenty-one rs-fMRI signal values and 28 radiomics features were selected to construct models. The performance of the combined models was exceptional, with the standout random forest (RF) model delivering an AUC value of 0.902 [95% confidence interval (CI): 0.813-0.990] in the validation set and 1.000 (95% CI: 1.000-1.000) in the training set. Further analysis of the 49 features revealed significantly overlapping brain regions for both feature types. Three key features demonstrating significant differences between ANI and PWND were identified (all P values <0.001). These features correlated with cognitive test performance (r>0.3). The RF combined model exhibited high classification performance in ANI, enabling objective and reliable individual diagnosis in clinical practice. It thus represents a novel method for characterizing the brain functional impairment and pathophysiology of patients with ANI. Greater attention should be paid to the frontoparietal and striatum in the research and clinical work related to ANI.