Prediction occurred an average of 75.8 months before final diagnosis.
A deep learning algorithm with fluorine 18 fluorodeoxyglucose PET of the brain improves early prediction of Alzheimer’s disease, according to a study published in the journal Radiology.
Researchers from the University of California in San Francisco sought to develop and validate a deep learning algorithm that predicts the final diagnosis of Alzheimer disease, mild cognitive impairment, or neither at fluorine 18 (18F) fluorodeoxyglucose (FDG) PET of the brain and compare its performance to that of radiologic readers.
Related article: Artificial intelligence in radiology: Friend or foe?
The researchers obtained 2,109 prospective 18F-FDG PET brain images from 1,002 patients from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and 40 imaging studies from 40 patients for the retrospective independent test set. Final clinical diagnosis at follow-up was recorded. Researchers trained the deep learning algorithm on 90% of the dataset and then tested it on the remaining 10% of the dataset. Through deep learning, the algorithm was able to teach itself metabolic patterns that corresponded to Alzheimer’s disease.
“Differences in the pattern of glucose uptake in the brain are very subtle and diffuse,” co-author Jae Ho Sohn, MD, said in a release. “People are good at finding specific biomarkers of disease, but metabolic changes represent a more global and subtle process.”
The algorithm achieved area under the ROC curve of 0.98 when evaluated on predicting the final clinical diagnosis of Alzheimer’s disease in the independent test set (82% specificity at 100% sensitivity), an average of 75.8 months prior to the final diagnosis, which in ROC space outperformed reader performance (57% [four of seven] sensitivity, 91% [30 of 33] specificity). Saliency map demonstrated attention to known areas of interest but with focus on the entire brain.
“We were very pleased with the algorithm’s performance,” Sohn said in the release. “It was able to predict every single case that advanced to Alzheimer’s disease.”
The researchers did caution that the study was small and needed further validation, but they stated that the algorithm could be a useful tool to complement the work of radiologists, especially in conjunction with other biochemical and imaging tests, in providing an opportunity for early therapeutic intervention.
Related article: AI and the Future of Radiology
“If we diagnose Alzheimer’s disease when all the symptoms have manifested, the brain volume loss is so significant that it’s too late to intervene,” Sohn said. “If we can detect it earlier, that’s an opportunity for investigators to potentially find better ways to slow down or even halt the disease process.”