Skip to content

Latest commit

 

History

History
51 lines (34 loc) · 1.94 KB

File metadata and controls

51 lines (34 loc) · 1.94 KB

Back to Projects List

Automated CT diaphragm muscle segementation

Key Investigators

  • Adamo Donovan (McGill University)
  • Dr. Benjamin McDonald Smith (McGill University)
  • Investigator 3 (Affiliation)

Project Description

Manual diaphragm segmentation requires 2 hours per left hemi-diaphragm for a trained rater.

Previous automated methods of CT diaphragm segmentation, have relied on using a priori anatomical knowledge (i.e. the lower surface of the lungs as a diaphragmatic landmark), mathematical models, and other assumptions in order to segment the diaphragm, which was then compared to the gold standard manual segmentation performed by expert radiologists. This method outputed measures of diaphragm distances, surface area, and curvature.

Our method would be the first to use the gold standard method of over 300 manual and direct diaphragmatic segmentations to train an artificial intelligence and neural network that would create a three-dimensional model of the left hemi-diaphragm with measures of density, volume, and dome height.

Objective

  1. Objective A. To automate diaphragm muscle segmentation from chest CT images using 300 manually segmented diaphragms along with an artificial neural network approach.

Approach and Plan

  1. Describe planned approach to reach objectives.
  2. ...
  3. ...

Progress and Next Steps

Illustrations

Background and References