Each year, ten million new cases of dementia are identified, but diagnosing different types of dementia and overlapping symptoms can complicate treatment. Researchers have now created an AI tool capable of diagnosing ten different types of dementia, including vascular dementia, Lewy body dementia, and frontotemporal dementia, even when they co-occur.
The researchers have developed a Machine Learning (ML) model that accurately identifies specific dementia-causing pathologies using commonly collected clinical data like demographic information, medical history of the patient and family, medication use, neurological and neuropsychological exam scores, as well as neuroimaging data such as MRI scans. These research findings are published online in Nature Medicine.
The corresponding author, Vijaya B. Kolachalama, PhD, FAHA, an associate professor at Boston University Chobanian & Avedisian School of Medicine, stated, “Our AI tool enables differential diagnosis of dementia using routinely collected clinical data, showing its potential as a scalable diagnostic tool for Alzheimer’s disease (AD) and related dementias.” Kolachalama emphasized the importance of generating diagnoses from routine clinical data due to challenges in accessing gold-standard testing, especially in remote areas and urban healthcare centers.
During the study, the ML model was trained on data from over 50,000 individuals across nine global datasets and achieved an impressive area under the receiver operating characteristic (ROC) curve of 0.96 for differentiating dementia types. The ROC score ranges from 0 to 1, where 0.5 signifies random guessing and 1 represents perfect performance.
The researchers compared the performance of neurologists and neuro-radiologists working alone versus using the AI tool. They found that the AI tool increased the accuracy of neurologists by more than 26% across all ten types of dementia. In a test with 100 randomly selected cases, 12 neurologists provided diagnoses and confidence scores ranging from 0 to 100, which were then combined with the AI tool’s probability score to generate an AI-augmented neurologist score.
Kolachalama highlighted the potential of AI in supporting healthcare systems by aiding in early identification of disorders and improving patient management. Given the increasing prevalence of dementia cases, the researchers believe that this AI tool can facilitate precise diagnoses and meet the rising demand for targeted therapeutic interventions in dementia care.
This research project received support from various grants, including those from the Karen Toffler Charitable Trust, the National Institute on Aging’s Artificial Intelligence and Technology Collaboratories, the American Heart Association, Gates Ventures, and the National Institutes of Health.
