Classifying Phonotrauma Severity from Vocal Fold Images with Soft Ordinal Regression

2025-11-12

AI Summary

Current clinical assessment of vocal fold injury severity relies on subjective expert judgment, resulting in high costs and poor reproducibility. To address this, we propose a soft-label-based ordinal regression framework that models inter-annotator variability as probabilistic soft labels—representing the distribution of expert ratings—and integrates them into a deep neural network architecture. We design an ordinal loss function explicitly tailored for soft labels to capture annotation uncertainty. Compared with conventional hard-label classification and standard ordinal regression, our method significantly improves both predictive accuracy and uncertainty calibration. Evaluated on a public vocal fold image dataset, it achieves near-expert-level performance. This work delivers the first automated analysis tool for vocal fold pathology that simultaneously attains high diagnostic accuracy and well-calibrated uncertainty estimation—enabling scalable screening, longitudinal monitoring, and reliable clinical decision support.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Phonotrauma refers to vocal fold tissue damage resulting from exposure to forces during voicing. It occurs on a continuum from mild to severe, and treatment options can vary based on severity. Assessment of severity involves a clinician's expert judgment, which is costly and can vary widely in reliability. In this work, we present the first method for automatically classifying phonotrauma severity from vocal fold images. To account for the ordinal nature of the labels, we adopt a widely used ordinal regression framework. To account for label uncertainty, we propose a novel modification to ordinal regression loss functions that enables them to operate on soft labels reflecting annotator rating distributions. Our proposed soft ordinal regression method achieves predictive performance approaching that of clinical experts, while producing well-calibrated uncertainty estimates. By providing an automated tool for phonotrauma severity assessment, our work can enable large-scale studies of phonotrauma, ultimately leading to improved clinical understanding and patient care.

Problem

• Automating phonotrauma severity classification from vocal fold images
• Addressing label uncertainty in ordinal regression for medical assessment
• Reducing reliance on costly clinician judgments through automated analysis

Innovation

• Using soft ordinal regression for vocal fold images
• Modifying loss functions to handle label uncertainty
• Providing automated phonotrauma severity classification tool