Machine Learning Techniques for DME Classification on SD-OCT images

How to use the pipeline?

Pre-processing pipeline

The follwoing pre-processing routines were applied:

• BM3D denoising,
• Flattening,
• Cropping.

Data variables

In the file pipeline/feature-preprocessing/pipeline_preprocessing.m, you need to set the following variables:

• data_directory: this directory contains the orignal SD-OCT volume. The format used was .img.
• store_directory: this directory corresponds to the place where the resulting data will be stored. The format used was .mat.

Algorithm variables

The variables which are not indicated in the inital publication and that can be changed are:

• x_size, y_size, z_size: the original size of the SD-OCT volume. It is needed to open .img file.
• sigma: the estimate of the standard deviation for the BM3D denoising.
• h_over_rpe, h_under_rpe, width_crop: the different variables driving the cropping.

Run the pipeline

From the root directory, launch MATLAB and run:

>> run pipeline/feature-preprocessing/pipeline_preprocessing.m

Extraction pipeline

For this pipeline, the following features were extracted:

• HOG,
• LBP with radius set to 8, 16, and 24. Both rotation and non-rotation invariant features were extracted.

Data variables

In the file pipeline/feature-extraction/pipeline_extraction_***.m, you need to set the following variables:

• data_directory: this directory contains the pre-processed SD-OCT volume. The format used was .mat.
• store_directory: this directory corresponds to the place where the resulting data will be stored. The format used was .mat.

Run the pipeline

From the root directory, launch MATLAB and run:

>> run pipeline/feature-extraction/pipeline_extraction_hog.m
>> run pipeline/feature-extraction/pipeline_extraction_lbp_8_ri.m
>> run pipeline/feature-extraction/pipeline_extraction_lbp_16_ri.m
>> run pipeline/feature-extraction/pipeline_extraction_lbp_24_ri.m
>> run pipeline/feature-extraction/pipeline_extraction_lbp_8_nri.m
>> run pipeline/feature-extraction/pipeline_extraction_lbp_16_nri.m
>> run pipeline/feature-extraction/pipeline_extraction_lbp_24_nri.m

Classification pipeline

The method for classification used was:

• Random Forest,
• Linear SVM,
• RBF-SVM.

Data variables

In the file pipeline/feature-classifier/_***.m, you need to set the following variables:

• data_directory: this directory contains the feature extracted from the SD-OCT volumes. The format used was .mat.
• store_directory: this directory corresponds to the place where the resulting data will be stored. The format used was .mat.
• gt_file: this is the file containing the label for each volume. You will have to make your own strategy.

Experiment 1

Classification of individual features

Move to the folder pipeline/feature-classification/experiment-1/individual/

The classifier name needs to be set with either linear_svm, rbf_svm, or random_forest.

From the root directory, launch MATLAB and run:

>> classifier_name = 'random_forest';
>> pipeline_classification_hog(classifier_name)
>> pipeline_classification_lbp_8_ri(classifier_name)
>> pipeline_classification_lbp_16_ri(classifier_name)
>> pipeline_classification_lbp_24_ri(classifier_name)
>> pipeline_classification_hog_pca(classifier_name)
>> pipeline_classification_lbp_8_ri_pca(classifier_name)
>> pipeline_classification_lbp_16_ri_pca(classifier_name)
>> pipeline_classification_lbp_24_ri_pca(classifier_name)

Classification of combined features

Move to the folder pipeline/feature-classification/experiment-1/combination/

The classifier name needs to be set with either linear_svm, rbf_svm, or random_forest.

From the root directory, launch MATLAB and run:

>> classifier_name = 'random_forest';
>> pipeline_classification_hog_lbp_8_ri_pca(classifier_name)
>> pipeline_classification_hog_lbp_16_ri_pca(classifier_name)
>> pipeline_classification_hog_lbp_24_ri_pca(classifier_name)

Experiment 2

Move to the folder pipeline/feature-classification/experiment-2/

The classifier name needs to be set with either linear_svm, rbf_svm, or random_forest. The number of visual words need also to be set.

From the root directory, launch MATLAB and run:

>> nb_words = 30;
>> classifier_name = 'random_forest';
>> pipeline_classification_hog_pca(classifier_name, nb_words)
>> pipeline_classification_lbp_8_ri_pca(classifier_name, nb_words)
>> pipeline_classification_lbp_16_ri_pca(classifier_name, nb_words)
>> pipeline_classification_lbp_24_ri_pca(classifier_name, nb_words)

Run the pipeline

From the root directory, launch MATLAB and run:

>> run pipeline/feature-classification/pipeline_classifier_pca_hog.m
>> run pipeline/feature-classification/pipeline_classifier_pca_hog_lbp_8_16_24_ri.m
>> run pipeline/feature-classification/pipeline_classifier_pca_hog_lbp_8_16_24_nri.m

Validation pipeline

Data variables

In the file pipeline/feature-validation/pipeline_validation_***.m, you need to set the following variables:

• data_directory: this directory contains the validation results. The format used was .mat.
• gt_file: this is the file containing the label for each volume. You will have to make your own strategy.

Experiment 1

Validation of individual features

Move to the folder pipeline/feature-validation/experiment-1/individual/

The classifier name needs to be set with either linear_svm, rbf_svm, or random_forest.

From the root directory, launch MATLAB and run:

>> classifier_name = 'random_forest'
>> pipeline_validation_hog(classifier_name)
>> pipeline_validation_lbp_8_ri(classifier_name)
>> pipeline_validation_lbp_16_ri(classifier_name)
>> pipeline_validation_lbp_24_ri(classifier_name)
>> pipeline_validation_hog_pca(classifier_name)
>> pipeline_validation_lbp_8_ri_pca(classifier_name)
>> pipeline_validation_lbp_16_ri_pca(classifier_name)
>> pipeline_validation_lbp_24_ri_pca(classifier_name)

Validation of combined features

Move to the folder pipeline/feature-validation/experiment-1/combination/

The classifier name needs to be set with either linear_svm, rbf_svm, or random_forest.

From the root directory, launch MATLAB and run:

>> classifier_name = 'random_forest'
>> pipeline_validation_hog_lbp_8_ri_pca(classifier_name)
>> pipeline_validation_hog_lbp_16_ri_pca(classifier_name)
>> pipeline_validation_hog_lbp_24_ri_pca(classifier_name)

Experiment 2

Move to the folder pipeline/feature-validation/experiment-2/

The classifier name needs to be set with either linear_svm, rbf_svm, or random_forest.

From the root directory, launch MATLAB and run:

>> classifier_name = 'random_forest'
>> pipeline_validation_hog_pca(classifier_name)
>> pipeline_validation_lbp_8_ri_pca(classifier_name)
>> pipeline_validation_lbp_16_ri_pca(classifier_name)
>> pipeline_validation_lbp_24_ri_pca(classifier_name)