Summary: Using neuroimaging data, a new deep learning algorithm was able to detect Alzheimer’s disease with 90.2% accuracy.
Source: mass general
Although researchers have made progress in detecting signs of Alzheimer’s disease using high-quality brain imaging tests collected through research studies, a team from Massachusetts General Hospital (MGH ) recently developed an accurate detection method based on routinely collected clinical brain images. This advance could lead to more accurate diagnoses.
For the study published in PLOS ONEMatthew Leming, PhD, a researcher at MGH’s Center for Systems Biology and researcher at Massachusetts Alzheimer’s Disease Research Center, and his colleagues used deep learning, a type of machine learning and artificial intelligence that uses large amounts of data and complex algorithms form patterns.
In this case, scientists developed a detection model for Alzheimer’s disease based on brain magnetic resonance image (MRI) data collected from patients with and without Alzheimer’s disease who were seen at the MGH before 2019.
Then the group tested the model on five datasets – MGH after 2019, Brigham and Women’s Hospital before and after 2019, and outside systems before and after 2019 – to see if it could accurately detect Alzheimer’s disease on the actual database. global clinical data, regardless of hospital and time.
Overall, the search involved 11,103 images from 2,348 patients at risk for Alzheimer’s disease and 26,892 images from 8,456 patients without Alzheimer’s disease. In all five datasets, the model detected Alzheimer’s disease risk with 90.2% accuracy.
Among the main innovations of this work was its ability to detect Alzheimer’s disease independently of other variables, such as age. “Alzheimer’s disease typically occurs in older people, so deep learning models often struggle to detect the rarest early cases,” says Leming.
“We solved this problem by making the deep learning model ‘blind’ to features of the brain that it finds too associated with the patient’s indicated age.”
Leming notes that another common challenge in disease detection, especially in the real world, is dealing with data that is very different from the training set. For example, a deep learning model trained on MRIs from a scanner manufactured by General Electric may not recognize MRIs collected on a scanner manufactured by Siemens.
The model used an uncertainty metric to determine if the patient data was too different from what it was trained on to make a successful prediction.
“This is one of the only studies to have used routinely collected brain MRIs to try to detect dementia. Although a large number of deep learning studies for the detection of Alzheimer’s disease from brain MRIs have been conducted, this study has taken substantial steps towards actually achieving this in clinical settings. real as opposed to perfect lab settings,” Leming said.
“Our results, with cross-site, cross-temporal, and cross-population generalizability, provide a strong case for the clinical use of this diagnostic technology.”
Additional co-authors include Sudeshna Das, PhD and Hyungsoon Im, PhD.
Funding: This work was supported by the National Institutes of Health and the Technology Innovation Program funded by the Ministry of Commerce, Industry and Energy of the Republic of Korea, managed through a sub- contract with the MGH.
About this research news on artificial intelligence and Alzheimer’s disease
Author: Bradon Chase
Source: mass general
Contact: Bradon Chase – Mass General
Picture: Image is in public domain
See also
Original research: Free access.
“Contradictory Regression and Uncertainty Measures for Classifying Heterogeneous Clinical MRI in Mass General Brigham” by Matthew Leming et al. PLOS ONE
Abstract
Conflicting Regression and Confounding Uncertainty Measures for Classifying Heterogeneous Clinical MRI in Mass General Brigham
In this work, we introduce a new deep learning architecture, MUCRAN (Multi-Confound Regression Adversarial Network), to train a deep learning model on clinical brain MRI while regressing demographic and technical confounders. .
We trained MUCRAN using 17,076 clinical T1 axial brain MRI scans collected from Massachusetts General Hospital before 2019 and demonstrated that MUCRAN can successfully regress major confounders in the large clinical dataset. We also applied an uncertainty quantification method on a set of these models to automatically exclude out-of-distribution data in AD detection.
By combining MUCRAN and the uncertainty quantification method, we showed consistent and significant increases in AD detection accuracy for newly collected MGH data (after 2019; 84.6% with MUCRAN versus 72.5% without MUCRAN) and for data from other hospitals (90.3% from Brigham and Women’s Hospital and 81.0% from other hospitals).
MUCRAN offers a generalizable approach for deep learning-based disease detection in heterogeneous clinical data.