462. Leveraging White Matter Microstructural Features to Construct a High-performance Diagnostic Classifier for Clinical High-risk for Psychosis
Year: 2025
Type: conference-abstract
Abstract: Background The early identification of individuals at clinical high risk for psychosis (CHR) is of paramount importance for implementing timely interventions and preventing the progression to full-blown psychotic disorders. Consequently, the development of a biologically grounded model for clinical diagnosis is essential. Extensive research has documented white matter microstructural abnormalities in CHR. Aims & Objectives This study aims to comprehensively explore whole-brain white matter microstructural information in individuals at CHR and to construct a high-performance diagnostic classifier based on such biological data. Method This study enrolled 40 CHR individuals and 54 demographically matched healthy controls (HCs). Diffusion magnetic resonance imaging (dMRI) and T1-weighted structural images were acquired. Using the Automated Fiber-tract Quantification (AFQ) method, diffusion index values were calculated for 100 equidistant nodes along 20 major long-distance fiber tracts. Additionally, a probabilistic fiber tracking approach was employed to quantify the connection probabilities and diffusion index values of fiber tracts within the frontal-striatal-thalamic circuits. All diffusion index values and connection probabilities were aggregated as initial features, and an optimal feature subset was selected using recursive feature elimination. A binary classifier distinguishing CHR from HC groups was trained using the random forest (RF) algorithm. Results In the training set, the binary classifier achieved an accuracy of 0.99, a sensitivity of 0.97, a specificity of 1 and an area under the curve (AUC) of 1.0. In the testing set, the accuracy was 0.68, a sensitivity of 0.6, a specificity of 0.78 and an AUC of 0.71. Key features contributing to classification included diffusion index values of the left anterior thalamic radiation, the right cingulate hippocampal tract and the fiber tract connecting the left nucleus accumbens to the left orbital gyrus. Discussion & Conclusions Microstructural information derived from dMRI-based fiber tract analysis is instrumental in developing a high-performance diagnostic classifier for CHR. Specific fiber tracts, such as the left anterior thalamic radiation, the right cingulate hippocampal tract, and the fiber tract connecting the left nucleus accumbens to the left orbital gyrus, demonstrate significant contributions to the classifier and may play a critical role in the neuropathology of CHR.
Institutions: Capital Medical University, Beijing Anding Hospital
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,693
OpenAlex ID: W4413284819