Explainable Alzheimer's Disease Detection Using linguistic features From Automatic Speech Recognition.

INTRODUCTION: Alzheimer's disease (AD) is the most prevalent type of dementia and can cause abnormal cognitive function and pro-gressive loss of essential life skills. Early screening is thus necessary for the preven-tion and intervention of AD. Speech dysfunc-tion is an early-onset symptom of AD pa-tients. Recent studies have demonstrated the promise of automated acoustic assessment using acoustic or linguistic features extracted from speech. However, most previous stud-ies have relied on manual transcription of text to extract linguistic features, which weakens the efficiency of automated as-sessment. The present study thus investi-gates the effectiveness of automatic speech recognition (ASR) in building an end-to-end automated speech analysis model for AD detection.

METHODS: We implemented three publicly available ASR engines and compared the classification performance using the ADReSS-IS2020 dataset. Besides, the SHapley Additive exPlanations (SHAP) algorithm was then used to identify critical features that contributed most to model performance.

RESULTS: Three automatic transcription tools obtained mean word error rate (WER) texts of 32%, 43%, and 40%, respectively. These automated texts achieved similar or even better results than manual texts in model performance for detecting dementia, achiev-ing classification accuracies of 89.58%, 83.33%, and 81.25%, respectively.

CONCLUSION: Our best model, using ensemble learning, is comparable to the state-of-art manual tran-scription-based methods, suggesting the possibility of an end-to-end medical assis-tance system for AD detection with ASR en-gines. Moreover, the critical linguistic fea-tures might provide insight into further stud-ies on the mechanism of AD.