• different classification models such as the MobileNet and the ResNet

What it does

  • classify the Genant Score which is used to grade the severity of vertebral fractures

How we built it

  • ResNet-34 pretrained on ImageNet adapted for classification of the Genant Score
  • Drawing images with equal probability for training

Challenges we ran into

  • class imbalance

Accomplishments that we're proud of

  • Mean-Squared-Error of 0.78 over the test dataset

What we learned

  • the impact of using pretrained models on medical tasks

What's next for suprised_by_deadline

  • happy to discuss results and experience with the other teams

Built With

Share this project: