New explainable AI model detects tremors from time-domain data

Researchers have developed a novel two-stage hierarchical framework for detecting tremors directly from time-domain kinematic data. This approach utilizes a combination of deep convolutional and long short-term memory networks, followed by a vision transformer, to learn and classify tremor patterns across entire trials. While achieving respectable F1-scores ranging from 0.594 to 0.947 across different body parts, it did not surpass the current state-of-the-art frequency-domain methods. However, the framework offers the advantage of minimal preprocessing and provides explainability through attention weights and Grad-CAM, highlighting temporal and anatomical tremor patterns. AI

IMPACT Offers a new data-driven approach to tremor detection with built-in explainability, potentially reducing reliance on expert-engineered features.

RANK_REASON The cluster contains an academic paper detailing a new AI methodology for tremor detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Deep convolutional and long short-term memory network
• Essential tremor
• Grad-CAM
• Parkinson's disease
• Vision transformer

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →