[ad_1]
Alzheimer's disease affects more than 5 million Americans, and is expected to reach 16 million by 2050. Early intervention can help alleviate the most debilitating symptoms, namely memory loss and the problems of reading and organizing thought. Unfortunately, this intervention remains one of the most difficult neurological disorders to detect. There is no specific test for Alzheimer's.
However, there may be hope on the horizon. Researchers from the Department of Radiology and Biomedical Imaging at the University of California and the Big Data in Radiology Group (BDRAD, a multidisciplinary team of physicians and engineers studying the science of radiological data) describe in a recently published study an AI system to predict Alzheimer's disease scans of the brain.
Their article (An In-depth Learning Model to Predict the Diagnosis of Alzheimer's Disease Using PET Scans at 18F-FDG ") was published in the journal Radiology this week.
"The differences in the configuration of glucose uptake in the brain are very subtle and diffuse," said Dr. Jae Ho Sohn, co-author. "People are good at finding specific biomarkers for the disease, but metabolic changes are a more global and subtle process."
Scientists at Unlearn.AI, a start-up that designs software tools for clinical research, have developed a system this year that predicts its progress. However, the UC Berkeley team chose to focus on a chemical marker that had not been used before to drive an AI model.
They developed a deep learning algorithm (stratified mathematical functions that mimic the behavior of human neurons) on positron emission tomography with 18-F-fluorodeoxyglucose (FDG-PET), a specialized imaging technique in which Patients receive an injection of FDG, a radioactive glucose compound, which allows radiologists – and in this case, an AI system – to measure the uptake into brain cells through PET. . It is an indicator of metabolic activity.
The researchers assembled a corpus from the Alzheimer's Disease Neuroimaging Initiative (ADNI), a project database containing over 2100 FDG-PET brain images from 1,002 patients. (About 10% of the samples were set aside for validation.) After learning the dataset, the AI system could track the minute changes in glucose uptake in certain areas of the brain that would be difficult to detect under normal circumstances.
In tests involving 40 imaging exams performed on 40 patients, the AI system of researchers reached a sensitivity of 100% when detecting Alzheimer's disease, averaging more than six years before the final diagnosis.
"We are very pleased with the performance of the algorithm," said Dr. Sohn. "It has been possible to predict every case that has progressed to Alzheimer's disease."
The team cautioned that this was its beginning – the sample size was relatively small. But they think the system could complement the work of radiologists and serve as a basis for an AI capable of identifying patterns of beta amyloid and tau protein accumulation, abnormal protein clusters, and Other biological markers associated with Alzheimer's disease. .
Source link
