• Segmentation, landmark detection, and angle measurement for corrective osteotomy.
• CT scans of the lower limbs were used to segment out parts of pelvis, femur, and tibia bones.
• These segmentations are then used to detect the important landmarks required to measure relevant angles for corrective osteotomy.
• Trained and deployed UNet and nnUNet for bone segmentation, and angle measurement.
• Implemented code to convert 3d segmentation masks predicted by the models to 3D meshes and save these volumes in stl format.
• Set-up an deployment environment using Flask and Docker, to deploy the trained models for segmentation and angle measurement.
• This project was done at NAAMII.
• Collaborator
  • Manish Dhakal
• Supervisor
  • Taman Upadhaya, Ph.D.