Alzheimer’s disease is by far the most common dementia around the world, the 6th leading cause of death in US and the related costs are more than $1 trillion US worldwide. In the previous studies, machine learning and deep learning algorithms are shown to successfully differentiate the brain MRIs of Alzheimer’s disease from healthy state of brain.
What it does, How we built it
In this work, we introduce an advanced approach where a 3D CNN trained by 1500 brain MRIs from three different stages: healthy, patients with mild cognitive impairment, and dementia, using Python. Later, the pixelwise predictions of the algorithm are shown in 3D superimposing the actual MRI, using Unity, which helps to visualize the effected regions.
Challenges we ran into
Training huge datasets required an extensive amount of computer work.
Accomplishments that we are proud of
Such work allows to differentiate between the stages, show the brain regions affected most, promising better prediction of the severity and the course of the disease. This approach may also lead to early diagnosis of the disease which in turn will improve the quality of life, efficiency of the treatment and reduce costs.
What's next for 3D CNN Analysis of Brain MRIs and Visualization
Patients with Alzheimer's disease can experience different symptoms depending on the brain region affected most. The prediction of the algorithm might be useful to compare with the existing symptoms which will further indicate the potential of this work.