A team of researchers at the Boston University School of Medicine (BUSM) have developed a machine learning AI algorithm that could help to detect Alzheimer’s disease and other neurodegenerative diseases, which could allow treatments to be provided sooner to help manage the diseases.
Alzheimer’s disease is the leading cause of dementia worldwide and the sixth leading cause of death in adults in the United States. It has been estimated that 1 in 10 adults over the age of 65 has Alzheimer’s disease. While the disease is incredibly common, detecting and diagnosing the disease remains a challenge. It is often diagnosed well after the onset of the disease.
A combination of techniques is currently used to detect and diagnose Alzheimer’s disease, including neuropsychological tests, MRI scans, and patient histories; however, effective practices are variably applied, and lack the necessary specificity and sensitivity.
Artificial intelligence algorithms have been developed to help clinicians detect a range of diseases from medical images and other data. The algorithms are trained on vast amounts of data and are capable of learning and becoming better at diagnosis over time, learning from any mistakes that are made.
The AI algorithm created by the BUSM researchers can accurately predict Alzheimer’s risk and diagnose the disease. The algorithm was trained using MRI scans of the brains of patients diagnosed with Alzheimer’s disease and individuals with normal cognitive function from four national cohorts, along with tests to measure the level of cognitive impairment, and patient demographics such as age and gender.
Data from one of the cohorts was used to develop the algorithm to accurately predict Alzheimer’s risk. The algorithm was then used on other data cohorts and was shown to accurately predict the disease status.
“Not only can we accurately predict the risk of Alzheimer’s disease, but this algorithm can generate interpretable and intuitive visualizations of individual Alzheimer’s disease risk en route to an accurate diagnosis,” said senior study investigator Vijaya Kolachalama, PhD, assistant professor of medicine at BUSM. “Our framework linked a fully convolutional network, which constructs high-resolution maps of disease probability from local brain structure to a multilayer perceptron and generates precise, intuitive visualization of individual Alzheimer’s disease risk en route to accurate diagnosis.”
The algorithm was tested on a range of different data sets and an international team of expert neurologists were asked to diagnose the disease from the same data. The researchers report the algorithm performed better than an average neurologist and found that the predictions of the system were highly aligned with autopsy reports on the brains of patients who had died.
If the AI system was readily available, it could help to speed up the diagnosis of Alzheimer’s disease, which could assist clinical practice, such as memory clinics.
You can read more about the study in the paper – Development and validation of an interpretable deep learning framework for Alzheimer’s disease classification – which was recently published in the journal Brain. DOI: doi.org/10.1093/brain/awaa137