A Concept-based approach to Voice Disorder Detection

2025-07-23

AI Summary

This study addresses the lack of trustworthiness and interpretability in AI-based voice disorder diagnosis in clinical practice. We propose a concept-driven, interpretable AI framework comprising a Concept Bottleneck Model (CBM) and a Concept Embedding Model (CEM), which map low-level speech deep features onto a clinically meaningful, interpretable concept space—such as abnormal vocal fold vibration or insufficient respiratory support—enabling end-to-end traceable decision-making. Our approach achieves diagnostic performance on par with black-box deep models (e.g., AUC ≥ 0.92) while, for the first time in speech pathology, delivering fine-grained, clinician-aligned, concept-level attributions. By explicitly grounding predictions in domain-specific medical concepts, the framework significantly enhances model transparency and clinician trust. It establishes a novel paradigm for deploying trustworthy AI in otolaryngology and speech-language pathology, bridging the gap between high-performing deep learning and clinical usability.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Voice disorders affect a significant portion of the population, and the ability to diagnose them using automated, non-invasive techniques would represent a substantial advancement in healthcare, improving the quality of life of patients. Recent studies have demonstrated that artificial intelligence models, particularly Deep Neural Networks (DNNs), can effectively address this task. However, due to their complexity, the decision-making process of such models often remain opaque, limiting their trustworthiness in clinical contexts. This paper investigates an alternative approach based on Explainable AI (XAI), a field that aims to improve the interpretability of DNNs by providing different forms of explanations. Specifically, this works focuses on concept-based models such as Concept Bottleneck Model (CBM) and Concept Embedding Model (CEM) and how they can achieve performance comparable to traditional deep learning methods, while offering a more transparent and interpretable decision framework.

Problem

• Detect voice disorders using automated non-invasive techniques
• Improve interpretability of AI models in clinical diagnostics
• Compare concept-based models with traditional deep learning methods

Innovation

• Concept-based models for voice disorder detection
• Explainable AI enhances DNN interpretability
• CBM and CEM balance performance and transparency