A groundbreaking multi-task learning framework, DEMENTIA, has been developed by Prof. LI Hai and his team at the Hefei Institutes of Physical Science of the Chinese Academy of Sciences, to improve the early detection and assessment of Alzheimer's disease (AD). 

The research was recently published in the IEEE Journal of Biomedical and Health Informatics.

As the global population ages, AD is becoming increasingly prevalent, making early detection critical for improving patient outcomes. Language decline is often one of the earliest indicators of cognitive decline. While automated speech analysis offers a non-invasive and cost-effective approach to detecting AD, existing methods face significant challenges, including complexity, poor interpretability, and limited integration of diverse data types, which hinder accuracy and clinical applicability.

To overcome these limitations, Prof. LI Hai's team developed the DEMENTIA framework. This innovative approach integrates speech, text, and expert knowledge using a hybrid attention mechanism, significantly enhancing both the accuracy and clinical interpretability of AD detection. 

Leveraging advanced large language model technologies, the framework captures intricate intra- and inter-modal interactions, improving detection accuracy and enabling the prediction of cognitive function scores. 

Furthermore, comprehensive interpretability analyses demonstrated the model's robust clinical decision-support capabilities and its adaptability across diverse datasets.

The findings underscore the potential of speech-based tools for early AD screening and monitoring cognitive decline. By providing a more accurate and interpretable solution, the DEMENTIA framework holds significant promise for addressing the challenges posed by an aging population, offering both scientific and societal value.

A Multi-Task Learning Framework for AD Assessment: Hybrid Attention and Multimodal Representation (Image by YANG Lizhuang)