lidaboo / octsegmentation Goto Github PK

This package provides a toolbox for segmenting OCT retina scans written in Matlab and C. A detailed documentation may be found in documentation.pdf. The corresponding paper can be found here: http://www.sciencedirect.com/science/article/pii/S1361841514000449.

1. Unzip the code into any directory

2. In Matlab set the bash variable OCT_CODE_DIR to the directory from step 1) via

setenv('OCT_CODE_DIR',xxx);

where xxx is a string pointing to that directory

3. call function compileMex to compile all C-Functions

The package provides two models (datafiles/modelFiles) for circular scans and 3-D volumes, that where trained using our labeled ground truth. Unfortunately we are not allow to publish any of these scans. For 2-D we have one demoscan provided, for 3-D we have demo scripts that use an external dataset.

1. run useExampleScan2D.m to obtain a segmentation for the circular scan that is included in the package
2. run octGUI: Load the matfile circularScan.mat (datafiles/exampleScans) and the model file (datafiles/modelFiles) and run the model by clicking on 'Segment'

1. Download the dataset of Srinivasan et al. from http://people.duke.edu/~sf59/Srinivasan_BOE_2014_dataset.htm
2. Run the script predSrinivasan.m for a demonstration how to segment volumes (uses the model provided in the package)
3. Run the script trainSrinivasan.m for a demonstration how to train a model for volumes using as ground truth the predictions from 2)

If you use this software in your publication, please cite the following publication:

"Probabilistic Intra-Retinal Layer Segmentation in 3-D OCT Images Using Global Shape Regularization" F Rathke, S Schmidt, C Schnörr, Medical Image Analysis

The application is free to use for research purposes.