an accurate diagnosis based on a single brain MRI

United Kingdom – A machine learning algorithm machine learning can determine whether a person has Alzheimer’s disease (AD) based on a single MRI with 98% accuracy, according to a study published online June 20 at Communications Medicine [1].

“Currently, no other simple and widely available method can diagnose Alzheimer’s disease with this level of accuracy, so our research is an important step forward,” the ministry said in a statement. Mr. Eric Aboagye from Imperial College London, which conducted the research.

“Many Alzheimer’s patients who attend memory consultations also have other neurological disorders, but even within this group our system can distinguish patients with Alzheimer’s from those without,” he adds.

A robust and reproducible tool

To develop the algorithm, Professor Aboagye and his colleagues divided the brain into 115 regions and assigned them 660 different characteristics, such as size, shape and texture. They trained the algorithm to identify where changes in this or that characteristic might accurately correspond to Alzheimer’s disease.

Using data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), the team tested their algorithm on brain MRIs from more than 400 patients with late-, early-, or advanced-stage Alzheimer’s disease, on healthy controls, and on patients with other conditions. neurological diseases, including frontotemporal. dementia and Parkinson’s disease.

They also tested it using data from more than 80 patients who underwent AD diagnostic testing at Imperial College Healthcare NHS Trust.

In 98% of cases, the MRI-based machine learning tool alone could accurately predict whether a person had Alzheimer’s disease, outperforming conventional measurements of hippocampal volume and hippocampal beta-amyloid protein in cerebrospinal fluid (CSF). It could also distinguish between early and advanced stages of Alzheimer’s disease in 79% of patients.

The tool was found to be “robust and reproducible across MRIs, demonstrating its potential for application in future clinical practice,” the researchers write.

“Most patients have to go through a full battery of tests before receiving a diagnosis and this tool could allow for faster diagnosis and reduce patient anxiety. Of course, the specialist could use this information to refine and modify the diagnosis,” said Professor Aboagye.

The algorithm also detected changes in brain areas not previously associated with Alzheimer’s disease, including the cerebellum and ventral diencephalon. This “opens up possibilities for researchers” to take a closer look at these areas and see how they may be related to dementia, Professor Aboagye said.

“Although neuroradiologists already interpret MRIs to help diagnose Alzheimer’s disease, some features of the scans may not be visible even to specialists,” explained Dr. Mr Paresh Malhotraco-investigator (Imperial College London) in the press release.

“Using an algorithm that can identify the texture and subtle structural features of the brain that are affected by Alzheimer’s disease could really improve the insights we can gain from standard imaging techniques,” Professor Malhotra added.

Need to replicate the experience.

speaking for Medscape Medical Newsyou Dr. Cyrus A. Rajiassistant professor of radiology and neurology at Washington University in St. Louis, Missouri, said the study shows that “new computational analyzes of T1-weighted or structural images can identify Alzheimer’s disease with a high degree of accuracy.”

However, “transposition into clinical practice will require replication of these results, as well as software optimized for the clinical setting,” Dr. Raji concluded.

This research was partially funded by the NIHR Center for Biomedical Research at Imperial College London and the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Prs Aboagye, Malhotra and Raji reported no relevant financial relationships.

The article was originally published on under the title Can a single brain scan accurately diagnose Alzheimer’s? Translated by Aude Lecrubier.

Follow Medscape in French at Twitter.

Follow | Medscape Cardiology at Twitter.

To receive our specialized newsletters.

Leave a Comment