DECT: Harnessing LLM-assisted Fine-Grained Linguistic Knowledge and Label-Switched and Label-Preserved Data Generation for Diagnosis of Alzheimer's Disease

TL;DR

This paper addresses the challenge of early Alzheimer's disease detection from noisy spontaneous speech by introducing DECT, an LLM-assisted framework that distills fine-grained linguistic knowledge and cognitive-linguistic atoms from transcripts. DECT combines CL atom extraction, linguistic marker identification, and label-switched/label-preserved (LSLP) data generation to create a robust AMR representation and multi-taskly optimizes a detector with losses $\\mathcal{L}_{CLS}$ and $\\mathcal{L}_{syn}$. The approach achieves strong empirical gains on the ADReSSo dataset, with the GPT4o configuration reaching $Acc=90.48\%$ and $F1=88.32\%$, surpassing traditional baselines and other LLM setups. The work demonstrates the potential of integrating structured linguistic representations and synthetic data generated by large language models to improve speech-based AD diagnosis, with implications for scalable, low-cost cognitive screening in clinical settings. $

Abstract

Alzheimer's Disease (AD) is an irreversible neurodegenerative disease affecting 50 million people worldwide. Low-cost, accurate identification of key markers of AD is crucial for timely diagnosis and intervention. Language impairment is one of the earliest signs of cognitive decline, which can be used to discriminate AD patients from normal control individuals. Patient-interviewer dialogues may be used to detect such impairments, but they are often mixed with ambiguous, noisy, and irrelevant information, making the AD detection task difficult. Moreover, the limited availability of AD speech samples and variability in their speech styles pose significant challenges in developing robust speech-based AD detection models. To address these challenges, we propose DECT, a novel speech-based domain-specific approach leveraging large language models (LLMs) for fine-grained linguistic analysis and label-switched label-preserved data generation. Our study presents four novelties: We harness the summarizing capabilities of LLMs to identify and distill key Cognitive-Linguistic information from noisy speech transcripts, effectively filtering irrelevant information. We leverage the inherent linguistic knowledge of LLMs to extract linguistic markers from unstructured and heterogeneous audio transcripts. We exploit the compositional ability of LLMs to generate AD speech transcripts consisting of diverse linguistic patterns to overcome the speech data scarcity challenge and enhance the robustness of AD detection models. We use the augmented AD textual speech transcript dataset and a more fine-grained representation of AD textual speech transcript data to fine-tune the AD detection model. The results have shown that DECT demonstrates superior model performance with an 11% improvement in AD detection accuracy on the datasets from DementiaBank compared to the baselines.

Figures (3)

• Figure 1: DECT: An LLM-assisted framework that leverages the inherent linguistic knowledge of LLMs to process spontaneous speech transcript. DECT presents a novel integrated LLM-assisted AD detection approach to Distill linguistic markers, Extract Cognitive-Linguistic (CL) Atoms, Compose diverse speech transcript data, and ultimately Tune task-specific models based on LLM extracted linguistic markers and distilled CL atoms and generated data.
• Figure 2: Visualization of Newly Generated AD and NC Speech Transcript Data with MIMIC Strategy and LSLP Strategy.
• Figure 3: Embedding visualization: Comparison of AM Representation and Original Transcript using t-SNE for AD and CN groups