matlab-image-processing-toolbox-v2

MATLAB Image Processing Toolbox

Expert skill for medical image analysis using MATLAB's Image Processing Toolbox (IPT R2025a+). Focuses on medical imaging patterns, domain-specific pipelines, and non-obvious gotchas that go beyond standard IPT knowledge.

When to Use This Skill

• Medical imaging pipelines: MRI, CT/X-ray, microscopy, histology
• Modality-specific denoising (Rician for MRI, speckle for ultrasound)
• Multi-tissue segmentation with domain-specific thresholds
• Cell counting, fluorescence quantification, stain normalization
• DL segmentation post-processing with IPT morphological cleanup
• Data type gotchas causing silent errors in medical image processing

Read Before Coding

Task	Read This	Template Script
MRI preprocessing	knowledge/cards/medical-mri.md	scripts/template_mri_preprocessing.m
Cell counting	knowledge/cards/medical-microscopy.md	scripts/template_cell_counting.m
CT windowing	knowledge/cards/medical-mri.md	scripts/template_ct_windowing.m
Fluorescence quantification	knowledge/cards/medical-microscopy.md	scripts/template_fluorescence_quantification.m
Histology stain normalization	knowledge/cards/medical-microscopy.md	scripts/template_histology_stain_normalization.m
Tissue thresholding	knowledge/cards/segmentation-thresholding.md	scripts/template_adaptive_thresholding.m
Watershed for touching cells	knowledge/cards/medical-microscopy.md	scripts/template_watershed_segmentation.m
Morphological mask cleanup	knowledge/cards/morphology-binary.md	scripts/template_morphological_cleanup.m
DL segmentation + IPT	knowledge/cards/deep-learning-segmentation.md	—
Data type gotchas	knowledge/cards/data-types.md	—
Modality-specific denoising	knowledge/cards/filtering-denoising.md	—
Advanced regionprops	knowledge/cards/feature-regions.md	—
Large image processing	—	scripts/template_large_image_blockproc.m
Edge detection pipeline	—	scripts/template_edge_detection_pipeline.m

Critical Rules

Rule 1: imfilter Zero-Pads by Default (Creates Dark Borders)

% imfilter defaults to zero-padding — dark border artifacts!
bad = imfilter(I, h);              % Dark borders
good = imfilter(I, h, 'replicate'); % Fix

% imgaussfilt defaults to 'replicate' — safe as-is
filtered = imgaussfilt(I, 2);  % Already correct

Rule 2: graythresh Returns [0,1], Not Pixel Values

level = graythresh(img_uint8);  % Returns 0.45, NOT 115

% Use imbinarize instead (handles automatically)
bw = imbinarize(img_uint8);

Rule 3: double() vs im2double() — Silent Disaster

bad = double(img_uint8);     % Still [0,255], NOT [0,1]!
good = im2double(img_uint8); % Correctly scales to [0,1]

Rule 4: SE Radius Must Be Smaller Than Features You Keep

% SE radius ≈ half the feature size you want to AFFECT
se = strel('disk', 5);  % Removes features ~10px diameter
cleaned = imopen(bw, se);

% If SE radius > feature radius, features disappear!

Rule 5: blockproc BorderSize Must Match Filter Radius

fun = @(block) imgaussfilt(block.data, 2);
result = blockproc(huge_image, [512 512], fun, ...
    'BorderSize', [6 6], ...      % 3*sigma for Gaussian
    'TrimBorder', true, ...
    'UseParallel', true);

Rule 6: DICOM Normalization — Use mat2gray, Not im2double

% im2double assumes standard uint16 range; DICOM data may differ
img_norm = mat2gray(double(dicomread('scan.dcm')));

Template Scripts

Ready-to-use medical imaging pipelines in scripts/:

Script	Purpose
template_mri_preprocessing.m	MRI bias correction, denoising, brain extraction
template_ct_windowing.m	CT Hounsfield unit windowing for different tissues
template_cell_counting.m	Automated cell detection and counting pipeline
template_fluorescence_quantification.m	Multi-channel fluorescence analysis
template_histology_stain_normalization.m	H&E color deconvolution and normalization
template_adaptive_thresholding.m	Adaptive thresholding with sensitivity tuning
template_watershed_segmentation.m	Marker-controlled watershed for touching objects
template_morphological_cleanup.m	Standard binary mask cleanup pipeline
template_large_image_blockproc.m	Block processing for large images
template_edge_detection_pipeline.m	Multi-method edge detection comparison

Knowledge Cards

Medical Imaging (primary value):

• knowledge/cards/medical-mri.md — MRI preprocessing, tissue segmentation, volumetry
• knowledge/cards/medical-microscopy.md — Cell counting, fluorescence, H&E histology

Medical-Specific Patterns:

• knowledge/cards/filtering-denoising.md — Modality-specific denoising gotchas
• knowledge/cards/segmentation-thresholding.md — Tissue segmentation thresholding
• knowledge/cards/morphology-binary.md — Medical cleanup recipes
• knowledge/cards/feature-regions.md — Advanced regionprops for medical measurement

Integration:

• knowledge/cards/deep-learning-segmentation.md — DL + IPT post-processing patterns
• knowledge/cards/data-types.md — Type conversion gotchas for medical formats

Cross-Toolbox Integration

For wavelet-based image processing (multiresolution denoising, fusion), use Wavelet Toolbox functions (wavedec2, wdenoise2) for multiresolution denoising and image fusion.

% Example: Wavelet + IPT fusion for MRI denoising
denoised = wdenoise2(mri, 'DenoisingMethod', 'Bayes');  % Wavelet
cleaned = imopen(denoised, strel('disk', 2));           % IPT morphology
enhanced = adapthisteq(cleaned);                        % IPT contrast

---

Verified against MATLAB R2025b