Finding brain tumors in MRI scans
Comparing three neural networks for the Dallas Regional Science Fair.
3Models compared
~96%Best accuracy
ResNet18Top model
The question
Doctors use MRI scans to find brain tumors and figure out what type they are. I wanted to know which kind of AI image model does the best job of sorting tumor types automatically, since a good model could help doctors work faster.
What I did
- Trained and tested three well-known image recognition models: ResNet18, Inception V3, and VGG16.
- Compared how accurately each one classified tumor types in brain MRI images.
- Found that ResNet18 worked best, at about 96% accuracy.
Presenting it
My project had to pass a selection round to get into the fair, since some papers were turned down. I presented my poster to Ph.D. professors who judged the Computational Biology division and answered their questions about my methods.
What's next
I'd like to try combining all three models, which could make the predictions even more accurate. I'm planning a new AI project for this year's science fair.