Skip to content

API.md

Latest commit

 

History

History
333 lines (238 loc) · 6.69 KB

API.md

File metadata and controls

333 lines (238 loc) · 6.69 KB

🔬 API Reference

Core Functions

VesselSegment

Main function for retinal vessel segmentation.

result = VesselSegment(imageFile, varargin)

Parameters:

  • imageFile (string): Path to retinal fundus image
  • method (optional): Segmentation method ('rf', 'multiscale', 'hybrid')
  • threshold (optional): Threshold value for binary segmentation

Returns:

  • result (struct): Segmentation results with fields:
    • segmentation: Binary vessel mask
    • probability: Probability map
    • features: Extracted features
    • metrics: Performance metrics (if ground truth available)

multi_test

Batch processing function for multiple images.

results = multi_test(dataset, method)

Parameters:

  • dataset (string): Dataset name ('DRIVE', 'STARE', 'CHASEDB1')
  • method (string): Segmentation method

Returns:

  • results (cell): Array of segmentation results

Feature Extraction

extractFeature

Extract standard features from retinal images.

features = extractFeature(image, mask)

Parameters:

  • image (uint8): RGB retinal image
  • mask (optional): Region of interest mask

Returns:

  • features (double): Feature vector

extractFeatureH

Extract hierarchical features using Local Haar Patterns.

featuresH = extractFeatureH(image, patchSize)

Parameters:

  • image (uint8): RGB retinal image
  • patchSize (int): Size of local patches

Returns:

  • featuresH (double): Hierarchical feature descriptors

Classification Functions

trainRFC

Train Random Forest classifier for vessel segmentation.

model = trainRFC(trainingPath, groundTruthPath, options)

Parameters:

  • trainingPath (string): Path to training images
  • groundTruthPath (string): Path to ground truth masks
  • options (struct): Training options
    • numTrees: Number of trees (default: 100)
    • maxDepth: Maximum tree depth (default: 10)

Returns:

  • model (struct): Trained Random Forest model

testRFC

Test Random Forest classifier on new images.

predictions = testRFC(model, testPath)

Parameters:

  • model (struct): Trained Random Forest model
  • testPath (string): Path to test images

Returns:

  • predictions (cell): Segmentation predictions

Line Detection Functions

get_lineresponse

Multi-scale line detection for vessel enhancement.

response = get_lineresponse(image, scales, options)

Parameters:

  • image (double): Grayscale retinal image
  • scales (vector): Scale values for line detection
  • options (struct): Detection options

Returns:

  • response (double): Line response at multiple scales

get_linemask

Generate binary vessel mask from line response.

mask = get_linemask(response, threshold)

Parameters:

  • response (double): Line response image
  • threshold (double): Threshold value

Returns:

  • mask (logical): Binary vessel mask

Preprocessing Functions

standardize

Standardize retinal image intensities.

standardImage = standardize(image, method)

Parameters:

  • image (uint8): Input retinal image
  • method (string): Standardization method ('histogram', 'zscore')

Returns:

  • standardImage (double): Standardized image

noisefiltering

Apply noise filtering to retinal images.

filteredImage = noisefiltering(image, filterType)

Parameters:

  • image (uint8): Noisy retinal image
  • filterType (string): Filter type ('gaussian', 'median', 'bilateral')

Returns:

  • filteredImage (uint8): Filtered image

fakepad

Add padding to images for boundary handling.

paddedImage = fakepad(image, padSize, method)

Parameters:

  • image (numeric): Input image
  • padSize (int): Padding size
  • method (string): Padding method ('symmetric', 'replicate')

Returns:

  • paddedImage (numeric): Padded image

Evaluation Functions

accuracy_tesst

Calculate segmentation accuracy metrics.

metrics = accuracy_tesst(prediction, groundTruth, mask)

Parameters:

  • prediction (logical): Predicted vessel mask
  • groundTruth (logical): Ground truth vessel mask
  • mask (optional): Field of view mask

Returns:

  • metrics (struct): Accuracy metrics
    • accuracy: Overall accuracy
    • sensitivity: Sensitivity (recall)
    • specificity: Specificity
    • precision: Precision
    • f1Score: F1-score
    • auc: Area under ROC curve

Utility Functions

addPaths

Add all necessary paths for the project.

addPaths()

Description: Automatically adds all source code directories to MATLAB path.

Binary Descriptor Functions

create_binary

Create binary descriptors for vessel detection.

descriptors = create_binary(image, keypoints)

Parameters:

  • image (uint8): Retinal image
  • keypoints (struct): Detected keypoints

Returns:

  • descriptors (logical): Binary descriptors

create_binary_32

Create 32-bit binary descriptors.

descriptors32 = create_binary_32(image, keypoints)

Parameters:

  • image (uint8): Retinal image
  • keypoints (struct): Detected keypoints

Returns:

  • descriptors32 (uint32): 32-bit binary descriptors

create_descriptor

Create custom descriptors for vessel characterization.

descriptors = create_descriptor(image, method, parameters)

Parameters:

  • image (uint8): Retinal image
  • method (string): Descriptor type ('lbp', 'glcm', 'gabor')
  • parameters (struct): Method-specific parameters

Returns:

  • descriptors (double): Feature descriptors

Error Handling

All functions include comprehensive error checking:

try
    result = VesselSegment('invalid_path.jpg');
catch ME
    switch ME.identifier
        case 'VesselSegment:FileNotFound'
            fprintf('Error: Image file not found\n');
        case 'VesselSegment:InvalidFormat'
            fprintf('Error: Unsupported image format\n');
        otherwise
            rethrow(ME);
    end
end

Performance Considerations

Memory Usage

  • Large images may require processing in patches
  • Use clear command to free memory between operations
  • Monitor memory usage with memory command

Processing Time

  • Multi-scale operations are computationally intensive
  • Consider parallel processing for batch operations
  • Use appropriate image scaling for faster processing

Optimization Tips

% Efficient batch processing
parfor i = 1:length(imageList)
    results{i} = VesselSegment(imageList{i});
end

% Memory-efficient processing
for i = 1:length(imageList)
    result = VesselSegment(imageList{i});
    save(['result_' num2str(i) '.mat'], 'result');
    clear result; % Free memory
end