Skip to content

USAGE.md

Latest commit

 

History

History
179 lines (134 loc) · 4.17 KB

USAGE.md

File metadata and controls

179 lines (134 loc) · 4.17 KB
Raw

📖 Usage Guide

Basic Usage

1. Quick Start Example

% Add paths and setup environment
run('src/utils/addPaths.m');

% Run quick start demo
run('scripts/quickStart.m');

2. Single Image Segmentation

% Load and segment a single retinal image
imageFile = 'Images/RFC SET/DRIVE/test/01_test.tif';
result = VesselSegment(imageFile);

% Display results
figure;
imshow(result.segmentation);
title('Vessel Segmentation Result');

3. Batch Processing

% Process multiple images from a dataset
dataset = 'DRIVE'; % Options: 'DRIVE', 'STARE', 'CHASEDB1'
results = multi_test(dataset);

Advanced Usage

Training Custom Models

Random Forest Classifier

% Prepare training data
trainingPath = 'Images/RFC SET/DRIVE/train/';
groundTruthPath = 'Images/RFC SET/DRIVE/rfc_mask/';

% Train RFC model
model = trainRFC(trainingPath, groundTruthPath);

% Test the model
testPath = 'Images/RFC SET/DRIVE/test/';
results = testRFC(model, testPath);

Feature Extraction

% Extract features from retinal images
imageFile = 'path/to/retinal/image.tif';

% Standard feature extraction
features = extractFeature(imageFile);

% Hierarchical feature extraction
featuresH = extractFeatureH(imageFile);

% Create binary descriptors
binaryDesc = create_binary(imageFile);
binaryDesc32 = create_binary_32(imageFile);

Multi-scale Line Detection

% Perform unsupervised vessel segmentation
imageFile = 'Images/RFC SET/DRIVE/test/01_test.tif';

% Get line response at multiple scales
lineResponse = get_lineresponse(imageFile);

% Generate line mask
lineMask = get_linemask(lineResponse);

Preprocessing Options

% Noise filtering
filteredImage = noisefiltering(originalImage);

% Image standardization
standardImage = standardize(originalImage);

% Fake padding for boundary handling
paddedImage = fakepad(originalImage, padSize);

Performance Evaluation

Accuracy Assessment

% Evaluate segmentation accuracy
groundTruth = imread('path/to/ground/truth.png');
segmentation = imread('path/to/segmentation/result.png');

% Calculate accuracy metrics
metrics = accuracy_tesst(segmentation, groundTruth);

% Display results
fprintf('Accuracy: %.2f%%\n', metrics.accuracy);
fprintf('Sensitivity: %.2f%%\n', metrics.sensitivity);
fprintf('Specificity: %.2f%%\n', metrics.specificity);

Configuration Options

Dataset-Specific Parameters

DRIVE Dataset

  • Image Size: 584 × 565 pixels
  • Training Images: 20 images
  • Test Images: 20 images
  • File Format: TIFF

STARE Dataset

  • Image Size: 700 × 605 pixels
  • Training Images: 10 images
  • Test Images: 10 images
  • File Format: PPM

CHASE_DB1 Dataset

  • Image Size: 999 × 960 pixels
  • Training Images: 8 images
  • Test Images: 20 images
  • File Format: JPG

Output Formats

The segmentation results are saved in multiple formats:

  1. Binary Masks: .png format in rfc_mask/ directories
  2. Processed Images: Color-coded results in rfc_output/ directories
  3. Multi-scale Results: Intermediate results in multiscale_mask/ directories

Tips for Best Results

  1. Image Quality: Ensure input images are high-quality and properly centered
  2. Preprocessing: Apply appropriate noise filtering for better results
  3. Parameter Tuning: Adjust threshold values based on dataset characteristics
  4. Memory Management: Process large datasets in batches to avoid memory issues

Common Workflows

Research Workflow

% 1. Setup environment
run('src/utils/addPaths.m');

% 2. Prepare datasets
% Place datasets in appropriate directories

% 3. Train models
model = trainRFC('DRIVE');

% 4. Test and evaluate
results = testRFC(model, 'DRIVE');
metrics = accuracy_tesst(results.segmentation, results.groundTruth);

% 5. Analyze results
% Generate plots and statistics

Clinical Application Workflow

% 1. Load clinical image
clinicalImage = 'path/to/clinical/fundus/image.tif';

% 2. Preprocess
processedImage = noisefiltering(imread(clinicalImage));

% 3. Segment vessels
segmentation = VesselSegment(processedImage);

% 4. Post-process and analyze
% Apply clinical analysis tools