AI tool improves diagnostic accuracy for dementia by 26%

AI tool can help diagnose dementia: ©Feng Yu - stock.adobe.com

Boston University researchers have developed an artificial intelligence tool designed to assist physicians in diagnosing the specific causes of cognitive decline. The findings were reported in Nature Medicine. While Alzheimer’s disease is the most recognized cause of dementia, it’s not the only one. The diagnostic challenge is compounded by the fact that multiple causes of dementia can occur simultaneously, making it difficult for physicians to arrive at a definitive diagnosis quickly. This delay often hampers timely intervention.

The researchers, led by Vijaya B. Kolachalama, PhD, an expert in using computational tools to aid in medical diagnoses, created an AI-driven platform capable of identifying up to 10 types of dementia, including vascular and frontotemporal dementia. This advanced tool integrates commonly collected patient data—such as medical history, medication use, demographic information, and scores from neurological and neuropsychological exams—with neuroimaging data like MRI scans. The AI then generates a prediction of the type of dementia a patient has, along with a confidence score, offering valuable insights to guide clinical decisions.

“Our goal is for AI to assist in identifying these disorders early, thereby enabling physicians to manage their patients more effectively and potentially prevent the diseases from worsening,” says Kolachalama, who serves as an associate professor of medicine and computer science at BU, in a statement.

The platform’s development represents a collaboration between BU researchers and external experts. Trained on data from over 50,000 individuals across nine global datasets, the AI tool has been rigorously tested, according to the researchers. In a study comparing neurologists working alone to those assisted by the AI, the tool improved diagnostic accuracy by 26%.

This AI tool is particularly valuable because it can function with limited data, which is crucial for health care providers in resource-constrained settings. Kolachalama said that in low-income regions, where MRI machines are less accessible, having a tool that can operate effectively with available clinical data is essential for expanding the reach of this technology.

Kolachalama also notes the increasing strain on health care systems due to a global shortage of neurology experts and a growing number of patients with neurological conditions. By enhancing diagnostic accuracy and efficiency, this AI tool has the potential to significantly alleviate the burden on physicians with limited time and resources.

Looking ahead, Kolachalama said he and his team are focused on bringing this AI platform into hospitals and clinics for real-world testing. The hope is that this technology will soon become an integral part of the diagnostic process, helping to improve outcomes for patients with dementia worldwide.