ichseg

ichseg is an automated tool for segmenting intracerebral hemorrhages (ICH) from CT scans. Key points:

- Uses machine learning, specifically a random forest classifier, to predict the probability of ICH at each voxel in a CT scan
- Trained and validated on 112 CT scans from the MISTIE clinical trial, which had manual ICH segmentations by experts
- Extracts imaging features from CT scans to use as predictors in the model
- Compared against logistic regression, LASSO, and generalized additive models
The random forest model achieved the highest median Dice Similarity Index of 0.899 when comparing automated segmentations to manual ground truth.
- Enables fast, automated estimation of hemorrhage location and volume, overcoming limitations of time-consuming manual segmentation.

Topic

Neurology;MRI;Neurobiology

Detail

  • Operation: Image analysis

  • Software interface: Command-line interface

  • Language: R

  • License: GNU General Public License, version 2

  • Cost: Free of charge with restrictions

  • Version name: v0.9.7

  • Credit: NIH, the National Institute on Aging, the National Institute of Neurological Disorders and Stroke, the National Institute of Mental Health.

  • Input: -

  • Output: -

  • Contact: John Muschelli jmusche1@jhu.edu

  • Collection: -

  • Maturity: -

Publications

  • PItcHPERFeCT: Primary Intracranial Hemorrhage Probability Estimation using Random Forests on CT.
  • Muschelli J, et al. PItcHPERFeCT: Primary Intracranial Hemorrhage Probability Estimation using Random Forests on CT. PItcHPERFeCT: Primary Intracranial Hemorrhage Probability Estimation using Random Forests on CT. 2017; 14:379-390. doi: 10.1016/j.nicl.2017.02.007
  • https://doi.org/10.1016/j.nicl.2017.02.007
  • PMID: 28275541
  • PMC: PMC5328741

Download and documentation

< Back to DB search