Although researchers have made considerable progress in using part of the collected high-quality brain imaging tests to detect the signs of Alzheimer's disease, a team from Massachusetts General Hospital (MGH) recently developed a more accurate detection method, which relies on clinical brain imaging collected routinely.
For this research published on PLOS ONE, Dr. Matthew Leming, a researcher at MGH Systems Biology Center and Alzheimer's Disease Research Center in Massachusetts, and his colleagues used deep learning, that is, machine learning and artificial use of a large amount of data and complex algorithms to train AI models.
Researchers developed an Alzheimer's disease detection model based on brain magnetic resonance imaging (MRI) data collected from patients with and without Alzheimer's disease who visited Massachusetts General Hospital before 2019.
Next, the team tested the model on five data sets to see whether it can accurately detect Alzheimer's disease based on real data. Structure. In all five data sets, the accuracy of the model in detecting the risk of Alzheimer's disease is 90.2%.
One of the main innovations of this work is that it can detect Alzheimer's disease without being affected by other variables, such as age.
"Alzheimer's disease usually occurs in the elderly, so the deep learning model is usually difficult to detect more rare cases of early onset," Dr. Leming said.
Leming pointed out that another common challenge of disease detection (especially in the real environment) is to process data that is different from the training set. For example, the deep learning model trained on the MRI of the scanner made by GE may not recognize the MRI collected on the scanner made by Siemens.
The model uses uncertainty measurement to determine whether the patient data is too different from the training data to make a successful prediction.
"This is one of the few studies that use routinely collected brain MRI images to detect dementia. Although a large number of in-depth learning studies have been conducted to detect Alzheimer's disease from brain MRI, this study has taken a substantial step towards the practical implementation of this goal in the real world clinical environment," Leming said. "Our results - with cross-site, cross-time and cross-population universality - provide a strong case for the clinical application of this diagnostic technology."