Inspiration
Increased life expectancy has led to a marked increase of the older population around the world. With this expanding older adult population, the US Alzheimer's disease burden is expected to nearly triple by the year 2060. Alzheimer's disease is a slowly progressing disease marked by changes in the structure and functioning of the brain, leading to declines in cognitive performance. Thus, early detection and intervention are of paramount importance for improving cognitive abilities in later life. This in turn can potentially be used to develop future machine learning approaches to other neurodegenerative diseases.
What it does
To assist with early detection of cognitive impairment, our team constructed a convolutional neural network to analyze structural and functional brain scans to assess declining brain regions and overall risk of cognitive impairment.
How we built it
To create this model, we utilized data from a phase-III clinical trial of 379 older adults, aged 65-89. Older adults were assessed with gold-standard neuropsychological tests, structural/functional MRI scans, and demographic information by a neuropsychologist to determine overall cognitive status.
Accomplishments that we're proud of
Through this process, we have achieved over 80% predictive accuracy to potentially serve as an early detection tool to detect and intervene in the progression of dementia. Further, our tool can be utilized for an individualized determination of brain regions that are predictive of cognitive impairment.
What we learned
This competition brought many challenges but our most important lesson was managing our time and learning to work effectively as a team. When it comes to time management, one of the approaches we could have implemented was to survey the various approaches that have been shown to work with this type of image data. We tuned our hyper parameters in a short period of time but we were unable to test transfer learning as a possible option for this problem.
What's next for brAIn Gang
While the model is only a prototype, we plan to develop the model further to be used in conjunction with cognitive intervention methods to improve outcomes in older adults.
Built With
- monai
- python
- pytoch
Log in or sign up for Devpost to join the conversation.