Every 3.2 seconds another person gets dementia! The most common form of dementia, up to 70%, is Alzheimer’s disease (AD). AD is a neurodegenerative disease causing the death of brain neurons, which commonly do not reproduce or get replaced. Depending on the level of AD, patients gradually begin to experience memory loss, impaired decision-making, language problems, loss of the ability
to think and function independently. Medication can positively affect patients with mild AD, while moderate and severe cases refuse treatment. Therefore, preventing damage of brain cells is an effective way to reduce the deterioration of the brain. We need to detect AD in the early stages to prevent death of brain cells before it is too late.
Researchers have detected AD by analyzing clinical data, genetics data and neuroimaging like magnetic resonance imaging (commonly known as MRI). They have developed methods to extract more information about how the brain changes by asking people for regular examinations over the course of multiple years, otherwise called a longitudinal study. These new methods could help to improve the prediction of AD and its treatment.
Neuroimaging is a method that can visualize changes in the brain structure and functions that cause AD symptoms. We have used longitudinal MRI data for analyzing the dynamics of brain volume changes to study the progression of AD. Our data includes patients with mild cognitive impairment, which is the step before getting AD, and patients who begin with mild cognitive impairment and progress to AD in later visits. We use this data to suggest a model to predict if people with mild cognitive impairment are at risk of AD. Our model measures the rate of change in the volume of different regions of an individual brain. It then applies machine learning on this information to diagnose the patient whose rate of brain change is abnormal. We tested and evaluated our model on ~5000 MRI scans of ~1000 people*.
The results show our method can potentially warn about AD earlier than clinical diagnoses. The model is able to distinguish stable and risky mild cognitive impairment before it is too late. This is one step further to increase the distances between new cases of dementia appearing every 3.2 seconds.
* Data are selected from ADNI (Alzheimer’s Disease Neuroimaging Initiative) data-set, which is a historic study of brain aging that started in 2004 and consists of a repository of data from North America and Canada that includes Alzheimer patients, mild cognitive impairment and elderly healthy controls.