Massachusetts General Hospital researchers are using automated MRI software to detect individuals in the preclinical phase of Alzheimer's disease with 95% accuracy, according to a cohort study.
The sooner physicians can diagnosis Alzheimer's disease (AD), the better. Early diagnosis means better treatment for the patient and more options for care.
Traditionally, AD has been diagnosed through a neurologic exam, detailed medical history, and written tests, with neuroimaging being used to rule out other conditions like stroke or a brain tumor. This study shows the potential and feasibility of an automated, MRI-based software program to find and define diagnostic markers for AD.
Dr. Rahul Desikan, a researcher at MGH, and colleagues verified which regions of the brain are affected by AD and mild cognitive impairment (MCI), a transitional phase between normal aging and Alzheimer's.
Based on earlier pathological and imaging studies, the investigators confirmed that individuals with AD or MCI demonstrate a significant difference in thickness and volume in their entorhinal cortex, hippocampus, and supramarginal gyrus (Brain 2009 May 21 epub ahead of print).
Once the researchers verified which regions of the brain are affected, they processed MRI scans of individuals with AD and MCI using FreeSurfer, openly available software developed at MGH and the University of California, San Diego. The software can be found at surfer.nmr.mgh.harvard.edu.
In their first cohort, involving 97 subjects, the researchers differentiated controls from MCI patients with 91% accuracy.
In the second cohort, involving 216 individuals, the researchers distinguished MCI patients from normal controls with 95% accuracy. They identified Alzheimer's patients with 100% accuracy.
Currently Medicare has approved an FDG-PET protocol for imaging Alzheimer's, but Desikan said MRI can provide different information.
"The current Medicare-approved FDG-PET protocol is mainly helpful for differentiating between dementia subtypes, specifically, differentiating frontotemporal dementia from Alzheimer's disease. Our MRI-based protocol is mainly for identifying those individuals who are in the preclinical phase of Alzheimer's disease," Desikan said.
Using MRI techniques and not just FDG-PET to image Alzheimer's patients could have repercussions for research and treatment, according to Desikan. "These techniques have powerful implications for clinical trials, specifically in their use as a longitudinal biomarker that can assess and quantify the efficacy of treatments," he said.
Although more and more researchers are using FreeSurfer, the software is not quite ready to serve as a clinical diagnostic tool, according to Debra Fleischman, Ph.D., a professor in the neurological sciences department at Rush University Medical Center in Chicago.
"Because not all persons with MCI develop a clinical diagnosis of AD, these measures need to be tested within a longitudinal design, preferably in a large epidemiological cohort, to determine if they have predictive utility for the conversion from MCI to AD," she said.
Also, FreeSurfer requires operator intervention and up to 40 hours for postprocessing, which hinders the software's clinical utility, Fleischman said.
"Each scan needs to be operator-checked for surface accuracies so that nonbrain tissue can be removed before submission to automated processing. This is particularly important in regions of the medial temporal lobe known to be affected in MCI and AD," she said.
Desikan concedes FreeSurfer isn't ready for clinical use pending follow-up studies.
Those studies will need to cover how to predict which individuals with MCI will progress to Alzheimer's, differentiating AD from other neurodegenerative diseases, and assessments of how these measures do at early diagnosis. All the findings will eventually need validation against the current gold standard for diagnosis: postmortem examination of the brain.