Cross-Lingual Alzheimer's Disease Detection Based on Scale Criteria

Alzheimer's Disease (AD) has become one of the most common neurodegenerative diseases worldwide, affecting patients' vocabulary, grammar, semantic processing, and speech fluency. This paper proposes a cross-lingual Alzheimer's Disease detection method based on a scale-rule analysis. To address the issues of data dependency and poor interpretability of traditional data-driven methods, we use the Z-score standard to extract and analyze scores for various aspects such as patients' expressive ability and speech fluency in different language texts. Detection is then performed through downstream models. This method not only offers better interpretability since all the features are extracted from the rating scales but also achieves an accuracy of 79.17% on the ADReSS20 dataset, 71.83% on the ADReSSo dataset, and 71.74% on the ADReSS-M dataset, demonstrating its effectiveness.