Language modeling screens Parkinson's disease with self-reported questionnaires.

Quest2Dx, a language-modeling AI for health questionnaires, screens Parkinson’s disease with AUROCs of 0.977 on PPMI Online and 0.974 on Fox Insight.

• Why it matters: Early, low-cost PD detection is hard because diagnosis depends on motor symptoms and invasive procedures, and questionnaire-based models often fail when responses are missing.
• What they did: Quest2Dx encodes the meaning of each questionnaire item and answer using language modeling, enabling transfer across different questionnaires and improving robustness to missing responses.
• The result: On PPMI Online and Fox Insight, it outperformed existing methods and achieved cross-questionnaire AUROCs of 0.920 and 0.952, while also highlighting key question predictors.

AI summary of the abstract · may contain errors · not medical advice · read the paper.