2016 Winter Project Week/Projects/Cluster-Driven Lung Segmentation

From NAMIC Wiki
Jump to: navigation, search
Home < 2016 Winter Project Week < Projects < Cluster-Driven Lung Segmentation

Key Investigators

  • Vivek Narayan (Dana Farber)
  • Raúl San José (BWH)
  • Dan Blezek (Isomics, Inc.)
  • Steve Pieper(Isomics, Inc.)
  • Chintan Parmar (Dana Farber)

Project Description

Objective Approach and Plan Progress and Next Steps
  • Consolidation of Lung Nodule detection and segmentation algorithms into pipelines.
  • Iterative parameter-space testing of pipelines in cluster-computing environments.
  • Perform evaluations against expert-contoured segmentations.
  • Explore and optimize existing Lung Nodule detection/segmentation tools.
  • Configure and launch StarCluster Nodes on AWS to run pipelines.
  • Use LIDC Data for evaluation of pipelines (~1200+ CT Images with manual contours).
  • Evaluate Segmentation pipelines against manual contours using metrics such as SimpleITK filters, Dice, Hausdorff, and Radiomics.
  • Exploring the Lesion-Size Toolbox algorithms for Lung Nodule detection.
  • Configured StarCluster/AWS testing pipeline with time and cost estimates

References

Slides describing initial experiments