AI software helps rule out Alzheimer’s disease from scans
Using software that measures the size and volume of brain structures from magnetic resonance imaging scans can help determine whether a patient has cognitive impairment.
The new research, from the University of California, could have implications for enhancing clinician capabilities for diagnosing Alzheimer’s disease, a progressive neurodegenerative disorder that is the most common cause of memory loss in older people. While there are drugs that can delay its progression, there is no effective treatment or cure for it.
There are other causes of cognitive impairment, some of which are treatable, such as hypertension and alcohol use. However, as much as 20 percent of Alzheimer’s disease is misdiagnosed, causing incorrect or delayed treatment for many people.
The researchers, from the University of California in Los Angeles (UCLA) and San Francisco (UCSF) proposed that using quantitative volume analysis software to validate volumetric MRIs would help detect whether patients had brain atrophies associated with Alzheimer’s disease or with other disorders.
The computing program, called Neuroreader, is an FDA-cleared medical image processing software that computes the physical size and volume of different brain regions on the MRI scans. It automates the current manual process of identifying, labeling and quantifying the volume of brain structures by measuring the scans against a data base of MRIs of healthy people.
Low hippocampal volumes in the mesial temporal lobe are strongly correlated to Alzheimer’s disease. The software can compute hippocampal volumes in less than five minutes and volumes of the entire brain in less than 10 minutes.
The computing program has been adopted at UCLA, the University of Pittsburgh Medical Center, and the Medical College of Wisconsin, according to neuroradiologist Cyrus Raji, MD, a clinical fellow in the radiology department at UCSF and one of the study’s authors.
“The main advantage of the software is its ability to measure and track brain volumes as a key indicator of brain health and neurodegenerative disease,” he says.
The researchers used the software to read the MRI scans of 22 patients who had been referred to UCLA’s Cognitive Health Clinic for memory decline. Their ages ranged from 53 to 92 years old.
Of the 22 patients, only five (23 percent) showed brain shrinkage patterns characteristic of Alzheimer’s disease.
For instance, the computing program reviewed the MRIs over a four-year period of one patient with bipolar disorder and memory loss. While the software found that the hippocampus and temporal volumes were borderline low, the volume loss was not progressive over time and not characteristic of Alzheimer’s disease. The patient was instead treated with a change in medication and placed on a physical activity program.
“There are many studies on brain volume, but ours is the first we know of that generates preliminary data examining the question of identifying other causes of memory loss distinct from Alzheimer's in a clinical sample with a wider array of potential causes of memory loss,” says Raji.
The researchers’ next steps include adding cases, using the software longitudinally to track changes in brain volume over greater time periods, and applying the software to other types of images, such as amyloid PET scans.
“Brain MRI with quantitative volumetric analysis can add value to the diagnostic evaluation of patients with cognitive impairment by identifying patterns of atrophy from causes other than Alzheimer's disease, such as vascular disease and psychiatric disorders,” the authors conclude.