AI improves questionnaire-based Parkinson's screening while working around missing answers

Scientists at Rensselaer Polytechnic Institute (RPI) have developed an artificial intelligence (AI) model called Quest2Dx to improve screening for Parkinson's disease (PD) in primary care settings. According to the paper published in Nature Aging, the AI system can more accurately analyze lengthy health questionnaires compared to other models, helping to identify early signs of PD.

Parkinson's diagnosis relies on observation of motor symptoms and invasive procedures, but the disease may be present for years before physical symptoms appear. Early detection and treatment would be more feasible if AI could scan health questionnaires for signs of Parkinson's. The Quest2Dx AI model was trained on the Parkinson's Precision Medicine Initiative (PPMI) database and the Fox Insight Study questionnaire, which contain 187 and over 300 questions, respectively.

The model can work around missing responses, handle multiple questionnaires with different wording, and calculate the most relevant question-answer pairs for each patient. The researchers identified difficulty with common tasks and exposure to industrial solvents as potential predictors of Parkinson's. The AI model was also able to create a more targeted survey with the 30 most pertinent questions, achieving nearly the same accuracy as the full questionnaire.

This development could lead to simplified questionnaires for PD screening, enabling early diagnosis and timely treatment for patients as the population ages and the risk of Parkinson's increases with age.