A hesitant fuzzy–entropy–TOPSIS framework for acoustic risk stratification of Parkinson’s disease patients

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that frequently causes speech impairments, making voice-based assessment a promising tool for disease monitoring and clinical decision-making. Although numerous studies have employed acoustic features for PD detection and classification, relatively limited attention has been given to interpretable patient-level risk stratification under acoustic uncertainty. Unlike conventional classification-oriented approaches, this study aims to develop an uncertainty-aware multi-criteria framework for ranking disease severity and stratifying patient risk based on heterogeneous speech biomarkers. To achieve this objective, a hesitant fuzzy–entropy–TOPSIS framework is proposed for acoustic risk assessment in PD patients. First, patient-level acoustic indicators are constructed from longitudinal voice recordings and represented as hesitant fuzzy elements to capture measurement variability and uncertainty. Subsequently, entropy weighting is employed to determine the objective importance of acoustic dimensions, and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is applied to generate individualized risk rankings and risk categories. The proposed framework was evaluated using the Parkinson’s Telemonitoring dataset, involving 42 patients and 16 core acoustic features grouped into four clinically meaningful dimensions: frequency instability, amplitude instability, noise characteristics, and nonlinear speech complexity. The results identified substantial heterogeneity in acoustic risk profiles across patients. The proposed model successfully generated continuous risk rankings and five-level risk stratification outcomes. External validation demonstrated a statistically significant association between the acoustic risk index and clinical severity measures. Additional robustness analyses showed that the ranking structure remained highly stable after excluding the most extreme observation (Spearman’s ρ = 0.998; Kendall’s τ = 0.978) and under an alternative robust ideal-solution definition (Spearman’s ρ = 0.999; Kendall’s τ = 0.988), indicating that the framework was not unduly influenced by a single patient. The proposed hesitant fuzzy–entropy–TOPSIS framework provides an interpretable and uncertainty-aware approach for PD acoustic risk stratification. By complementing traditional classification-based methods with patient-level severity ranking and structured risk assessment, this study offers a practical decision-support tool for precision monitoring and individualized management of Parkinson’s disease.