lxmwust / choroidsegmentation Goto Github PK

Open-source algorithm for automatic choroid segmentation of OCT volume reconstructions
--------------------------------------------------------------------------------------

In this file you will find the description of a piece of software for the
automatic choroid segmentation of the OCT volume scans. This is the matlab
implementation of the algorith described in detail in the paper:

Open-source algorithm for automatic choroid segmentation of OCT volume reconstructions
Javier Mazzaferri, Luke Beaton, Gisèle Hounye, Diane N. Sayah, and Santiago Costantino, 
Scientific Reports 7. Article Number 42112, (2017). doi:10.1038/srep42112

Directory structure
-------------------

The software is organized in a set of functions (See section <Processing>) 
that have to be executed in a predefined order. The functions take 
as parameter a cell-array of strings, each one containing the absolute path
to the base directory (VOLUMEDIR) of a particular OCT volume scan. 

At start, each VOLUMEDIR must be populated with a directory labeled 
"RawImages", with the list of images containing all Bscans of the OCT volume.
An image of the fundus can be stored directly under VOLUMEDIR.

During the process the information is organized under this directory as in 
the following structure:

VOLUMEDIR/
    DataFiles: Intermediate data generated in the process
    ProcessedImages: Sorted set of source images in png format for each Bscan
    RawImages: Contains the raw Bscans (and the .xml file, for Spectralis *)   
    Results: The results of the segmentation in different formats

* If your data does not come from a Heilderberg Spectralis OCT, you will 
need to set manually some parameters. See the begining of the function 
prepareVolumeGeneric.    

Processing
----------

The user needs to set writing permission on VOLUMEDIR.

To build choroid maps, call the following functions in this order:

1 - prepareVolumeGeneric (or prepareVolumeSpectralis)
2 - trimDetails
3 - mapPseudoRegistration
4 - retinaLayersSegmentation
5 - choroidMap
6 - choroidMovie

Each function takes as parameters a cell array of strings, each containing 
the path of one experiment data. For example, to process data on
directories VOL1, VOL2, and VOL3, located under the /home directory execute
the following commands on the matlab command window:

>> addpath(genpath('CODEDIR'))
>> dataDirs = {'/home/VOL1','/home/VOL2','/home/VOL3'};
>> prepareVolumeGeneric(dataDirs);
>> trimDetails(dataDirs);
>> mapPseudoRegistration(dataDirs);
>> retinaLayersSegmentation(dataDirs);
>> choroidMap(dataDirs);
>> choroidMovie(dataDirs);

CODEDIR is the path to this matlab code.

Description of functions
------------------------
prepareVolumeGeneric: Generates the file imageList.mat containing 
metainformation of each B-scan. Some parameters has to be set manually if 
the OCT volume data does not come from a Heidelberg Spectralis OCT. 
If you hava data generated with a Heidelberg Spectralis OCT, export all
Bscans and the .xml file the RawImages folder, and run the function 
prepareVolumeSpectralis, instead.

trimDetails: This function is the only intective function. It allows the 
user to set the starting and ending columns withing the Bscans, where to 
start and end looking for the choroid. This is done to avoid parts of 
bscans where the choroid is not clearly visible.

mapPseudoRegistration: This function rearrages the Bscans to ease processing.

retinaLayersSegmentation: This does the segmentation of the retinal layers 
and the choroid. It stores the information in segmentationResults.mat.

choroidMap: creates the 2D map of the choroid. Stores the result in 
ChoroidMap.mat file, and generates the figure ChoroidMap.pdf. 

choroidMovie: creates an animated gif file, showing a sequence of all the 
segmented Bscans, along with the corresponding line in the 2D map.

The script loadParameters.m set several parameteres of the process.