3D Kidney Segmentation

Automated kidney structure segmentation from medical scans

Challenges

Manual segmentation of kidney structures from CT/MRI scans is time-intensive for radiologists, prone to human error, and limits the ability to provide quantitative analysis for diagnostic decision-making.

Solutions

3D convolutional neural networks using U-Net architecture variants with medical imaging libraries to automatically segment kidney structures from volumetric medical scans.

Key Results

  • Dice Score: 0.86
  • IoU: 0.84

Technology

Pytorch, 3D Segmentation, CT Images, FC-CNN

View on GitHub