Machine Learning Analysis of Posturography in Panic Disorder: A Pilot Study for Objective Physiological Biomarker Identification

Luiz Antonio Vesco Gaiotto^1*Felipe Oliveira de Aguiar^1,2Thales Marcon Almeida¹Lucas Murrins Marques^1,2RICARDO RIYOITI UCHIDA^1,2

• ¹Mental Health Department, Santa Casa de Sao Paulo School of Medical Sciences, São Paulo, Brazil, SAO PAULO, Brazil
• ²Infinity Doctor's Inc. Artificial Intelligence Research Division, DE, United States of America, Miami, United States

Panic disorder (PD) is linked to subtle abnormalities in postural control, inadequately captured by traditional statistics. Machine learning (ML) techniques applied to stabilometric data may enhance detection of PD-related postural patterns. Objective: Evaluate static postural control in PD patients and determine if ML analysis of multivariate stabilometric data can improve differentiation from healthy controls. Methods: In this cross-sectional case-control study, 12 adults diagnosed with DSM-5 PD and 21 matched healthy volunteers (total n=33; 341 force platform trials) underwent stabilometry under five sensory conditions. Classical statistics used repeated-measures ANOVA on baseline trials only (to preserve independence). ML models (Decision Tree, k-Nearest Neighbors, Linear Discriminant Analysis, Logistic Regression, Random Forest) were trained under stratified, subject-grouped 4-fold cross-validation (StratifiedGroupKFold), ensuring all trials from each participant were confined to a single fold to prevent leakage, for the explainability of the model, we accessed LIME (Local Interpretable Model-Agnostic Explanations). Results: ANOVA revealed a significant group and condition interaction for mediolateral CoP displacement (p<0.01), with PD patients exhibiting consistently reduced mediolateral sway. No significant between-group differences emerged for anteroposterior sway. Using an optimized decision threshold (Youden), Logistic Regression achieved accuracy 93.8% and AUC = 96%; Linear Discriminant Analysis presented the highest specificity (91.7%). Conclusions: This is the first study applying ML to posturography for identifying physiological markers of panic disorder, using ML for stabilometric data, improves classification accuracy, highlighting static posturography as superior as clinical screening tools like PHQ-PD and panic disorder severity scale . Larger, externally validated cohorts and portable measurement solutions are needed.