Deep learning model also identifies different types of cognitive impairment.
AI can diagnose Alzheimer’s disease and other brain diseases as accurately as doctors, according to a new research paper.
The paper, which was published in Nature Communications, found that using a deep learning model can identify various stages of cognitive impairment – normal, mild impairment, Alzheimer’s disease or non-Alzheimer’s dementias.
Importantly, researchers found that “these frameworks compare favorably with diagnostic accuracy of practicing neurologists and neuroradiologists.”
The findings highlight the potential of using AI to efficiently and accurately diagnose large swaths of the elderly as the number of aged population accelerates worldwide, increasing the need for health services.
With more elderly and a predicted shortage of doctors, the AI diagnostic tool can be especially helpful in rural areas where specialists are not as readily available. Also, it could bring therapies to patients at a faster pace since a diagnosis can be made more readily.
“Even in circumstances where a specialized neurologist or neuro-radiologist is busy to directly provide a diagnosis, it is foreseeable that some degree of automation could step in to help,” said study co-author Vijaya B. Kolachalama, assistant professor of medicine at Boston University School of Medicine.
Past research used simple computational models to detect only whether a patient had a neurological disease or not. But the practical world is more nuanced, with a team of doctors typically evaluating patients for several possible explanations for their dementia.
This AI model comes closer to using all the data that clinicians typically use to understand a patient’s disease and make an accurate diagnosis.
“We show that this is achievable when a model is presented with a broad differential diagnosis of possible illnesses. Our study is novel because, unlike work before it, we demonstrate a computational strategy for providing an accurate diagnosis during this diverse landscape of neurologic disease,” said Kolachalama.
The researchers developed an array of models, using specialized methods in machine learning to capture the vast amounts of data used in making a diagnosis. The information included evaluative data from physical examinations, laboratory results and neuro-psychological test scores.
They used the same methods to examine MRI scans and drilled down to localized regions to look for degenerative tissue changes.
The last phase of the study was a “head-to-head” comparative analysis, using the AI model versus an international group of clinicians. Both groups evaluated the same patient data. The accuracy of the diagnosis between the doctors and the AI model was similar.