TL;DR
Turn quiet afternoons into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
Researchers at Rensselaer Polytechnic Institute developed Quest2Dx, an AI framework that analyzes health questionnaire responses to flag people who may need further Parkinson’s testing. In a study published in Nature Aging, the system outperformed comparison models and was nearly as accurate with a 30-question survey as with longer questionnaires; it is a screening tool, not a diagnostic system.
Researchers at Rensselaer Polytechnic Institute have developed an artificial intelligence framework that analyzes health questionnaires for patterns associated with Parkinson’s disease, potentially helping clinicians identify people who may warrant further testing. In a paper published in Nature Aging, the team reported that its Quest2Dx system outperformed earlier models and came close to its full-questionnaire performance using a shorter, 30-question survey. The researchers describe it as a screening approach, not a way to diagnose the disease.
Quest2Dx was developed by a team led by Pingkun Yan, head of RPI’s Department of Biomedical Engineering. The researchers trained the model using data from the Parkinson’s Progression Markers Initiative (PPMI) and tested it against a separate dataset from the Fox Insight study. The questionnaires in those datasets contain 187 and more than 300 items, respectively, and respondents may leave some questions unanswered.
The framework is designed to handle missing answers, recognize questions with different wording or order, and identify which combinations of questions and responses are most relevant for an individual. The authors reported that it performed better than earlier approaches, including one based on the pretrained language model Llama 2. They also created a survey containing 30 questions ranked as most informative by Quest2Dx; the model’s predictions with that shorter survey were nearly as accurate as with the full questionnaires.
Among the responses the system ranked as informative were difficulties with everyday activities such as getting out of bed or eating, along with reported exposure to industrial solvents and degreasers. The paper also discusses possible indicators such as sleep disorders, anxiety and changes in smell. The researchers caution that these factors can occur in other conditions and that no single questionnaire response establishes Parkinson’s disease.
Earlier Screening Through Shorter Surveys
Parkinson’s disease can be present for years before its characteristic movement symptoms become apparent, according to the researchers. People may not be referred to specialists until later, while changes such as disrupted sleep, loss of smell, handwriting difficulty or mood shifts may be dismissed as minor or attributed to other causes. A system that flags combinations of responses could help primary-care clinicians decide who may need specialist assessment or further testing.
The potential practical benefit is also in reducing the burden of long surveys. If a shorter questionnaire retains much of the model’s reported predictive performance, it could be easier for patients to complete and for clinics to use. That could support earlier referrals, but the study does not establish that using Quest2Dx improves patient outcomes or shortens the time to diagnosis in routine care.
Products Worth Considering
Fitbit Versa 2 Health and Fitness Smartwatch with Heart Rate, Alexa, Music, Sleep and Swim Tracking
Withings Sleep Under-Mattress Sleep Tracking Pad
Blueair Blue Pure 211i Max Air Purifier for Large Rooms
As an affiliate, we earn on qualifying purchases.
From Questionnaires to Clinical Referral
Parkinson’s diagnosis currently draws on clinical assessment of symptoms, including observation and measurement of motor changes, and may involve additional procedures. The RPI team is not proposing to replace that process. Its stated goal is to screen questionnaire responses for people who might benefit from further evaluation, potentially before more visible physical symptoms emerge.
Health questionnaires can be difficult for conventional machine-learning systems to analyze. Respondents may skip items, and standard approaches may fill in missing answers through imputation, which the researchers say can introduce errors. Models can also be sensitive to the order or phrasing of questions. Quest2Dx was designed to interpret the meaning and context of questions and answers, rather than rely only on their position in a form.
The study used PPMI data for model development and Fox Insight data for validation. The authors’ results are based on these research datasets; the supplied report does not describe a prospective trial in which primary-care clinics used the tool to guide referrals.
“We built this AI model to screen for potential cases earlier.”
— Pingkun Yan, RPI biomedical engineering department head
Clinical Accuracy Still Needs Testing
The reported results do not establish how Quest2Dx would perform among a broad, unselected population attending primary-care clinics. The source material does not provide the model’s specific accuracy figures, such as sensitivity, specificity or rates of false-positive and false-negative results, nor does it describe a prospective clinical evaluation. Those measures would help show how reliably the system distinguishes people who may have Parkinson’s from those with similar symptoms or risk factors.
It also remains unclear how the tool would fit into clinical workflows, how clinicians should act on a screening result, and whether a shorter survey would perform similarly across different patient groups. Questionnaire signals such as anxiety, sleep problems or solvent exposure can have multiple explanations. A positive screen would not confirm Parkinson’s and would require appropriate clinical follow-up.
Testing the Tool in Care Settings
The researchers say their findings could inform the design of simplified questionnaires for Parkinson’s screening and future use in research or primary care. The next step would be evaluation in real-world clinical settings, including testing performance across varied patient populations and measuring whether the screening process leads to useful referrals.
Until such evidence is available, Quest2Dx should be understood as a research framework that identifies patterns in questionnaire data. The paper’s reported results do not show that the system is already in routine use or that it can replace assessment by a qualified clinician.
Key Questions
Does Quest2Dx diagnose Parkinson’s disease?
No. The researchers describe it as a screening tool intended to flag people who may need further evaluation. A screening result is not a diagnosis.
How did the researchers evaluate the framework?
The team developed the model using data from the PPMI database and used responses from the Fox Insight study as a separate validation dataset. The paper compared Quest2Dx with earlier models.
Can it work with a shorter questionnaire?
The researchers reported that a survey containing the 30 questions Quest2Dx ranked as most informative produced predictions nearly as accurate as those using the full questionnaires. The source material does not give specific accuracy figures.
Do symptoms such as anxiety or sleep problems mean someone has Parkinson’s?
No. The study identifies these as possible patterns in questionnaire data, but they can also occur for many other reasons. No single response is diagnostic; concerns about symptoms should be discussed with a qualified healthcare professional.
Source: news-medical.net
