Deep Learning Analysis of Neuroimaging (MRI) Data
Public DepositedAttention Deficit Hyperactivity Disorder (ADHD), impacts the lives of a significant portion of people around the world. Individuals diagnosed with ADHD tend to have worse academic performance than their peers, are more prone to addiction, and have been found to have a generally lower quality of life. Given the negative impacts of this developmental disorder as well as it's prevalence, early and accurate diagnosis is very important. The current standard for ADHD diagnosis is the ADHD rating scale, or ADHD-RS, which is a self-reported questionnaire. In recent years, however, the usage of deep learning for diagnosis has become a topic of increasing interest. Particularly, numerous studies have been conducted to use deep learning to identify a diagnosis using Magnetic Resonance Imaging (MRI) scans. A deep learning network could potentially achieve higher accuracy in diagnosis than humans, especially with children, where the boundary between ADHD symptom and typical child-like behavior becomes blurred. We hoped to use a pretrained neural network to train a new network that could identify ADHD and it's subtypes from anatomical MRI scans. In doing so, we hoped to potentially gain a deeper insight into ADHD, how the brain is structured, as well as how effective it would be to use pretrained networks across different MRI classification. We found that our network was unable to improve using the anatomical MRI scans. This could indicate an ineffectiveness in using pretrained neural networks across different diagnosis. It could also indicate that anatomical MRI scans are not useful for identifying ADHD. Due to time constraints, we were unable to test these results further, however, we propose that further research into how neural networks pretrained on MRI data react to a change in classifications could provide important insight into brain function.
- This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.
- Creator
- Subject
- Publisher
- Identifier
- 119186
- E-project-032224-194256
- Advisor
- Year
- 2024
- UN Sustainable Development Goals
- Date created
- 2024-03-22
- Resource type
- Source
- E-project-032224-194256
- Rights statement
Relations
- In Collection:
Items
Items
Thumbnail | Title | Visibility | Embargo Release Date | Actions |
---|---|---|---|---|
|
Shields_MRI_IQP_Paper.pdf | Public | Download |
Permanent link to this page: https://digital.wpi.edu/show/ft848w075