Quantifying Articulatory Coordination as a Biomarker for Schizophrenia

TL;DR

This work addresses the interpretability gap in AI-based schizophrenia assessment by proposing a severity-sensitive, speech-based biomarker. It introduces an interpretable pipeline that extracts articulatory features via acoustic-to-articulatory inversion, builds full vocal tract coordination (FVTC) matrices, derives eigenspectra, and computes a weighted exponential-decay score (WSED) defined as $WSED = \\sum_{i=1}^{n} v_i \\alpha^{i-1}$ with $\\alpha=0.8$. Across 40-second segments from 23 individuals with schizophrenia and healthy controls, WSED distinguishes complex (negative) from simple (positive) coordination and correlates with overall BPRS severity and the Positive–Negative symptom balance. The results support a transparent, clinically interpretable biomarker that could inform severity-aware, speech-based assessments and guide personalized care.

Abstract

Advances in artificial intelligence (AI) and deep learning have improved diagnostic capabilities in healthcare, yet limited interpretability continues to hinder clinical adoption. Schizophrenia, a complex disorder with diverse symptoms including disorganized speech and social withdrawal, demands tools that capture symptom severity and provide clinically meaningful insights beyond binary diagnosis. Here, we present an interpretable framework that leverages articulatory speech features through eigenspectra difference plots and a weighted sum with exponential decay (WSED) to quantify vocal tract coordination. Eigenspectra plots effectively distinguished complex from simpler coordination patterns, and WSED scores reliably separated these groups, with ambiguity confined to a narrow range near zero. Importantly, WSED scores correlated not only with overall BPRS severity but also with the balance between positive and negative symptoms, reflecting more complex coordination in subjects with pronounced positive symptoms and the opposite trend for stronger negative symptoms. This approach offers a transparent, severity-sensitive biomarker for schizophrenia, advancing the potential for clinically interpretable speech-based assessment tools.

Figures (3)

• Figure 1: Eigenspectra difference plots of a subject with complex corrdination and a subject with a more simpler coordination
• Figure 2: Mean WSED plotted in relation with the BPRS scores of the Schizophrenia subjects
• Figure 3: Mean WSED plotted in relation with the Difference of positive and negative symptoms subscale scores