paper

From Black-Box to Clinical Insight: A Multi-Stage Explainable Framework for Speech-Based Cognitive Impairment Detection

paperactiveprovisional

from-black-box-to-clinical-insight-a-multi-stage-explainable-framework-for-speech-based-cognitive-impairment-detection-f11dccc8

·1 events·first seen 14h ago

Aliases: From Black-Box to Clinical Insight: A Multi-Stage Explainable Framework for Speech-Based Cognitive Impairment Detection

Co-occurring entities

NIA PREPARE Llama 3.3 70B Instruct SpeechCARE-Adaptive Gating Network SHapley Additive exPlanations Meta

More like this (12)

Compositional Reasoning Depth Predicts Clinical AI Failure Cognitive Episodes in LLM Reasoning Traces Enable Interpretable Human Item Difficulty Prediction Language-Based Digital Twins for Elderly Cognitive Assistance Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models Dialogue to Detection: A Multimodal Hybrid NLP Pipeline for Insurance Fraud Detection Expert-Aware Causal Tracing of Factual Recall in Sparse MoE Language Models Beyond task performance: Decoding bioacoustic embeddings with speech features Multilingual Reasoning Cascades Need More Context Measuring User's Mental Models of Speech Translation in Human-AI Collaboration Towards Explainable Adjudicative Variance: Quantifying Judicial Discretion via Gated Multi-Task Learning Modeling Complex Behaviors: Multi-Personality Composition and Dynamic Switching in Vision-Language Models Cross-Modal Masking for Robust Silent Speech Synthesis Using sEMG and Lipreading

Recent events (1)

4arXiv · cs.CL·14h ago·source ↗

Multi-stage explainability framework translates transformer speech models into clinical cognitive impairment narratives

A new arXiv preprint proposes a framework for making transformer-based speech cognitive impairment detection clinically interpretable by combining SHAP token attribution, linguistic feature analysis, and a four-stage LLM reasoning pipeline using LLaMA-3.1-70B-Instruct. The system is built on the SpeechCARE-Adaptive Gating Network multimodal model (F1=72.11% on NIA PREPARE) and maps outputs to four cognitive-linguistic dimensions. Physician evaluation on 70 samples showed strong alignment with clinical profiles and a System Usability Scale score of 82/100, suggesting practical clinical workflow integration potential.

Evaluation and Benchmarking AI Safety Research NIA PREPARE Llama 3.3 70B Instruct SpeechCARE-Adaptive Gating Network +3 more