Confidence Estimation for Automatic Detection of Depression and Alzheimer's Disease Based on Clinical Interviews
This work addresses the need for reliable diagnostic tools in healthcare by providing a method to reduce misdiagnosis risks, though it is incremental as it builds on existing speech-based detection approaches.
The paper tackled the problem of improving trust in automatic detection of Alzheimer's disease and depression from speech by proposing a novel Bayesian method for confidence estimation, which outperformed baselines in classification accuracy and confidence estimation on ADReSS and DAIC-WOZ datasets.
Speech-based automatic detection of Alzheimer's disease (AD) and depression has attracted increased attention. Confidence estimation is crucial for a trust-worthy automatic diagnostic system which informs the clinician about the confidence of model predictions and helps reduce the risk of misdiagnosis. This paper investigates confidence estimation for automatic detection of AD and depression based on clinical interviews. A novel Bayesian approach is proposed which uses a dynamic Dirichlet prior distribution to model the second-order probability of the predictive distribution. Experimental results on the publicly available ADReSS and DAIC-WOZ datasets demonstrate that the proposed method outperforms a range of baselines for both classification accuracy and confidence estimation.