New AI tool flags early risk of cognitive decline in Parkinson’s
Routine blood tests may help identify patients facing future memory problems
Written by |
- A new artificial intelligence model predicts cognitive decline and memory problems in Parkinson's disease patients using routine clinical data.
- Higher blood triglyceride levels and standard clinical factors like age are key risk indicators linked to future cognitive impairment.
- Further prospective multicenter validation is required before clinicians can implement this tool for routine patient monitoring.
An artificial intelligence model that analyzes routine checkup data and blood fat levels may help doctors spot which people with Parkinson’s disease are at higher risk for future memory and thinking problems.
By flagging at-risk individuals early, this approach could one day help doctors and families plan ahead long before nonmotor symptoms begin to interfere with daily life.
“These findings support further prospective, multicenter investigation of routinely available biomarkers for cognitive risk stratification and of the temporal relationship between [fat] metabolism and cognitive decline in [Parkinson’s],” researchers wrote.
The study, “Prediction of cognitive impairment in Parkinson’s disease using TabPFN: an integrative analysis of clinical data, longitudinal trajectories, and lipidomics,” was published in Frontiers in Aging Neuroscience.
Predicting cognitive changes in Parkinson’s
Cognitive impairment, such as difficulty with attention and memory, is a common nonmotor symptom of Parkinson’s. Often, people with Parkinson’s do not have signs of cognitive impairment when they are diagnosed, but then later on, as the disease progresses, they start to develop cognitive problems.
However, not everyone with Parkinson’s will develop cognitive impairment. Currently, clinicians don’t have reliable ways to predict whether an individual diagnosed with Parkinson’s is more or less likely to have cognitive problems in the future. In theory, a tool that reliably predicts the risk of future cognitive issues could help patients and their families make proactive plans to manage them when they arise.
In this study, scientists in China explored whether machine learning (a branch of artificial intelligence in which algorithms learn patterns directly from patient data) could predict future cognitive issues in Parkinson’s disease. Here, the researchers specifically used a machine learning framework called a Tabular Prior-data Fitted Network (TabPFN), an approach designed to analyze standard medical records and lab spreadsheets quickly and accurately.
To develop their model, the researchers trained the machine learning tool using data from 91 patients who received care at Beijing Tiantan Hospital between 2018 and 2025. None of these patients had cognitive impairment at initial assessment, but over several years of follow-up, about half of them progressed to meet criteria for cognitive impairment. The researchers trained their model using data routinely collected through clinical assessments and lab tests.
“From a clinical perspective, the use of routinely available variables may facilitate accessible risk stratification without requiring … specialized measurements,” the team added. “The model may therefore be useful for identifying patients who warrant closer cognitive monitoring.”
How blood fats influence the model
The researchers’ model showed overall good accuracy. In the original dataset, it accurately predicted approximately three-quarters of patients who developed cognitive impairment, and accurately identified a similar proportion of those who did not. The researchers then tested the model on an external dataset from the Parkinson’s Progression Markers Initiative, and it accurately identified 75% of patients with cognitive impairment and 68.9% of those without.
Notably, the final model included standard clinical factors such as age, education history, and initial cognitive scores, as well as blood triglyceride (TG) levels, a type of fat routinely measured during standard cholesterol checks. Data indicated that higher TG levels were associated with a greater likelihood of cognitive impairment, and in exploratory analyses in which TG data were removed, the model showed somewhat reduced accuracy, particularly in distinguishing patients who would not experience cognitive decline.
The researchers stressed that these associations don’t prove that higher blood fat levels lead to cognitive decline in Parkinson’s, but the data suggest that fat levels and cognition are somehow linked. As such, these data should be interpreted “as a predictive association that warrants further investigation rather than evidence that TG directly contributes to cognitive decline,” the researchers wrote.
More broadly, the scientists emphasized that these analyses were limited to a fairly small number of patients, underscoring the importance of further studies to verify and expand upon these results. “Prospective multicenter validation, assessment of calibration across clinical settings, and evaluation of clinical utility are required before implementation” of the tool, they wrote.
Leave a comment
Fill in the required fields to post. Your email address will not be published.