Medical Image Processing MATLAB Projects (22+ Ideas with Complete Code)

% 1. Read microfluidic microscope frame sequence
vidReader = VideoReader('microfluidic_cell_stream.avi');
kalman = vision.KalmanFilter('ConstantVelocity', 'StateTransitionModel', [1 1; 0 1]);

while hasFrame(vidReader)
    frame = readFrame(vidReader);
    gray = im2gray(frame);
    
    % 2. Detect spherical cells via circular Hough transform
    [centers, radii, metric] = imfindcircles(gray, [15 35], 'Sensitivity', 0.88);
    
    % 3. Real-time vision-based feedback trajectory error
    if ~isempty(centers)
        targetPos = [250, 200]; % Target coordinate in microchannel
        currentPos = centers(1, :);
        posError = targetPos - currentPos; % Control input to EK voltage
    end
end

% 1. Load microscopic blood smear image and transform to L*a*b* space
img = imread('blood_smear_wbc.jpg');
lab = rgb2lab(img);
a_channel = lab(:,:,2); % Nucleus appears strongly in high chromaticity

% 2. Segment Nucleus using Otsu threshold on 'a*' chrominance
level_nuc = graythresh(a_channel);
nuc_mask = imbinarize(a_channel, level_nuc);
nuc_mask = bwareaopen(nuc_mask, 100); % Remove small debris artifacts

% 3. Morphological extraction of entire WBC and cytoplasm subtraction
wbc_mask = imbinarize(rgb2gray(img), 0.7);
cytoplasm_mask = xor(wbc_mask, nuc_mask);

% 4. Calculate Nucleus-to-Cytoplasm (N:C) Ratio
nuc_area = sum(nuc_mask(:));
cyto_area = sum(cytoplasm_mask(:));
nc_ratio = nuc_area / (cyto_area + eps);

% 1. Establish Serial Stream with Microcontroller Node
s = serialport("COM4", 115200);
configureTerminator(s, "CR/LF");
flush(s);

% 2. Initialize Real-Time Visualization Window
hFig = figure('Name', 'Wireless Vital Signs Monitor');
hLinePulse = animatedline('Color', 'r', 'LineWidth', 1.5);
xlabel('Time (Samples)'); ylabel('PPG Amplitude'); grid on;

% 3. Stream and calculate instant heart rate
for k = 1:500
    dataStr = readline(s);
    sensorVals = str2double(split(dataStr, ",")); % [Temp, PPG]
    addpoints(hLinePulse, k, sensorVals(2));
    drawnow limitrate;
end

% 1. Compute AC and DC components from Red (660nm) & IR (940nm) PPG
fs = 100; % 100 Hz sampling
ppg_red_filt = bandpass(raw_red, [0.5 5], fs);
ppg_ir_filt  = bandpass(raw_ir, [0.5 5], fs);

ac_red = max(ppg_red_filt) - min(ppg_red_filt);
dc_red = mean(raw_red);
ac_ir  = max(ppg_ir_filt) - min(ppg_ir_filt);
dc_ir  = mean(raw_ir);

% 2. Compute Ratio-of-Ratios (R) and empirical SpO2 calibration
R = (ac_red / dc_red) / (ac_ir / dc_ir);
SpO2 = 110 - 25 * R; % Standard empirical linear calibration

fprintf('Calculated Arterial Oxygen Saturation: %.2f%%\n', SpO2);

% 1. Load Plate Well Image with Reagent Reaction (Anti-A, Anti-B, Anti-D)
wellImg = imread('well_anti_a.png');
grayWell = rgb2gray(wellImg);

% 2. Extract Texture GLCM Features for Agglutination Clumping
glcm = graycomatrix(grayWell, 'Offset', [0 1; -1 1; -1 0; -1 -1]);
stats = graycoprops(glcm, {'Contrast', 'Homogeneity', 'Energy'});

% 3. Classify reaction: High contrast & low homogeneity indicate clumping
isAgglutinated = mean(stats.Contrast) > 1.8 && mean(stats.Homogeneity) < 0.65;
if isAgglutinated
    disp('Anti-A Reaction: Positive (Agglutination Detected)');
else
    disp('Anti-A Reaction: Negative');
end

% 1. Envelope Detection of Beamformed RF Signal Matrix via Hilbert Transform
rf_env = abs(hilbert(rf_beamformed_data));

% 2. Dynamic Range Log Compression (60 dB clinical standard)
bmode_img = 20 * log10(rf_env / max(rf_env(:)));
bmode_img(bmode_img < -60) = -60;

% 3. Anisotropic Speckle Diffusion Filtering (Preserves cyst/organ edges)
bmode_clean = imdiffusefilt(mat2gray(bmode_img), 'NumberOfIterations', 8);

figure; imagesc(bmode_clean); colormap(gray); axis image off;
title('Speckle-Reduced B-Mode Ultrasound Scan');

% 1. Load ALL-IDB Microscopic Blood Cell Dataset
imds = imageDatastore('ALL_IDB_Dataset', 'IncludeSubfolders', true, 'LabelSource', 'foldernames');
[imdsTrain, imdsVal] = splitEachLabel(imds, 0.8, 'randomized');

% 2. Transfer Learning using Pretrained ResNet-18
net = resnet18;
lgraph = layerGraph(net);
newFc = fullyConnectedLayer(2, 'Name', 'fc_leukemia', 'WeightLearnRateFactor', 10);
newClass = classificationLayer('Name', 'output');
lgraph = replaceLayer(lgraph, 'fc1000', newFc);
lgraph = replaceLayer(lgraph, 'ClassificationLayer_predictions', newClass);

% 3. Train Classifier with Adam Optimizer
options = trainingOptions('adam', 'InitialLearnRate', 1e-4, 'MaxEpochs', 15, 'MiniBatchSize', 32, 'ValidationData', imdsVal, 'Plots', 'training-progress');
trainedNet = trainNetwork(imdsTrain, lgraph, options);

% 1. Load Brain MRI Slice and perform CLAHE enhancement
I = dicomread('brain_mri_flair.dcm');
I_norm = mat2gray(I);
I_enh = adapthisteq(I_norm, 'ClipLimit', 0.02);

% 2. Skull stripping via morphological closing & masking
mask_skull = imbinarize(I_enh, 0.1);
mask_skull = imfill(bwareafilt(mask_skull, 1), 'holes');
I_brain = I_enh .* mask_skull;

% 3. Chan-Vese Active Contour for hyperintense tumor region
mask_init = false(size(I_brain));
mask_init(120:180, 200:260) = true; % Seed ROI
seg_tumor = activecontour(I_brain, mask_init, 300, 'Chan-Vese');

figure; imshow(I_enh); hold on;
visboundaries(seg_tumor, 'Color', 'r', 'LineWidth', 2);
title('Delineated Brain Tumor Active Contour');

% 1. Extract Green Channel from Color Fundus Image (Highest Vessel Contrast)
fundus = imread('retina_drive_01.tif');
green_ch = fundus(:,:,2);
green_inv = imcomplement(green_ch);

% 2. Multidirectional 2D Gabor Filter Bank (12 Orientations)
wavelengths = [4 8];
orientations = 0:15:165;
[mag, phase] = imgaborfilt(green_inv, wavelengths, orientations);
vessel_response = max(mag, [], 3);

% 3. Adaptive Thresholding and Morphological Cleanup
bw_vessels = imbinarize(vessel_response, 'adaptive', 'Sensitivity', 0.55);
bw_clean = bwareaopen(bw_vessels, 30); % Remove isolated noise pixels

% 1. Load Pre-Trained DenseNet Model for Chest X-Ray Triage
net = densenet201;
inputSize = net.Layers(1).InputSize;

% 2. Predict on Test Chest Radiograph
testImg = imread('covid_chest_xray.png');
imgResized = imresize(testImg, inputSize(1:2));
[label, scores] = classify(trainedCovidNet, imgResized);

% 3. Compute Explainable Grad-CAM Activation Heatmap
camMap = gradCAM(trainedCovidNet, imgResized, label);
figure; imshow(imgResized); hold on;
imagesc(camMap, 'AlphaData', 0.45); colormap jet;
title(sprintf('Prediction: %s (Confidence: %.1f%%)', string(label), max(scores)*100));

% 1. Hair Removal via Morphological Black Top-Hat & Inpainting
lesion = imread('dermoscopy_isic_02.jpg');
se = strel('disk', 6);
hair_mask = imtophat(rgb2gray(lesion), se) > 0.08;
lesion_clean = inpaintCoherent(lesion, hair_mask);

% 2. Segment Lesion using Otsu on Blue Channel
bw = ~imbinarize(lesion_clean(:,:,3));
props = regionprops(bw, 'Area', 'Perimeter', 'Eccentricity', 'MajorAxisLength');

% 3. Compute Asymmetry and Border Compactness (Circularity)
circularity = (4 * pi * props.Area) / (props.Perimeter^2);
TDS = 1.3*props.Eccentricity + 0.1*(1/circularity) + 0.5*std2(lesion_clean(:,:,1));

% 1. Read Volumetric DICOM CT Series
[V, spatial] = dicomreadVolume('CT_Chest_Series_Dir');
V = squeeze(V); % 512x512xN Volumetric matrix

% 2. Extract Bone Isosurface (Hounsfield Threshold > 300 HU)
boneThreshold = 300;
fv = isosurface(V, boneThreshold);

% 3. Render 3D Anatomical Mesh with Phong Lighting
figure('Color', 'k');
p = patch(fv, 'FaceColor', [0.9 0.85 0.75], 'EdgeColor', 'none');
isonormals(V, p);
daspect(spatial.PixelSpacings);
view(3); axis tight; camlight; lighting phong;
title('3D Reconstructed Skeletal CT Mesh', 'Color', 'w');

% 1. 2D Discrete Wavelet Decomposition to isolate high-frequency calcifications
mammo = imread('mammogram_mias_03.pgm');
[LL, LH, HL, HH] = dwt2(double(mammo), 'bior3.5');

% 2. Morphological Top-Hat on detail sub-bands to emphasize spots
high_detail = LH + HL + HH;
microcalc_cand = imtophat(mat2gray(high_detail), strel('disk', 3));
bw_spots = microcalc_cand > 0.35;

% 3. Filter clusters by size and spatial density
clusters = bwpropfilt(bw_spots, 'Area', [3 50]);
figure; imshow(mammo, []); hold on;
visboundaries(imresize(clusters, 2), 'Color', 'y', 'LineWidth', 1.5);
title('Detected Microcalcification Clusters');

% 1. Binarize Cell Image & Compute Distance Transform
I = rgb2gray(imread('histology_cells.png'));
bw = imbinarize(I, 'adaptive');
bw_filled = imfill(bw, 'holes');
D = -bwdist(~bw_filled);

% 2. Marker-Controlled Watershed to split touching nuclei
mask = imextendedmin(D, 2);
D_mod = imimposemin(D, mask);
L = watershed(D_mod);
bw_split = bw_filled;
bw_split(L == 0) = 0; % Watershed ridge lines

% 3. Enumerate Connected Cell Regions
stats = regionprops(bw_split, 'Centroid', 'Area');
fprintf('Total Segmented Cells Counted: %d\n', length(stats));

% 1. Perform U-Net Semantic Segmentation across Cardiac Phases
pxEndo = semanticseg(cineSliceMatrix, trainedCardiacUNet);
lv_mask = (pxEndo == 'LeftVentricle');

% 2. Compute Volume over all short-axis slices (Simpson's Rule)
sliceThickness_mm = 8.0;
EDV = sum(lv_mask(:,:,:,phase_ED), 'all') * pixelArea * sliceThickness_mm * 1e-3; % mL
ESV = sum(lv_mask(:,:,:,phase_ES), 'all') * pixelArea * sliceThickness_mm * 1e-3; % mL

% 3. Compute Ejection Fraction (EF%)
EF = ((EDV - ESV) / EDV) * 100;
fprintf('Cardiac Metrics -> EDV: %.1f mL | ESV: %.1f mL | EF: %.1f%%\n', EDV, ESV, EF);

% 1. Load Registered Anatomical MRI (Gray) & Functional PET (Color)
mri = im2double(imread('mri_structural.png'));
pet = im2double(imread('pet_functional.png'));

% 2. Multilevel 2D Wavelet Fusion (Maximum Selection Rule for Detail Bands)
wname = 'sym4';
fused_img = wfusmat(mri, rgb2gray(pet), wname, 'max', 'mean');

% 3. Display Fused Diagnostic Visual
figure;
subplot(1,3,1); imshow(mri); title('Anatomical MRI');
subplot(1,3,2); imshow(pet); title('Functional PET');
subplot(1,3,3); imshow(fused_img); title('Wavelet Fused Composite');

% 1. Preprocess Orthopedic X-ray
xray = imread('femur_fracture.png');
xray_adj = imadjust(rgb2gray(xray));
bw_edges = edge(xray_adj, 'canny', [0.08 0.22]);

% 2. Hough Transform for Discontinuity & Line Tracking
[H, theta, rho] = hough(bw_edges);
peaks = houghpeaks(H, 5, 'threshold', ceil(0.3*max(H(:))));
lines = houghlines(bw_edges, theta, rho, peaks, 'FillGap', 10, 'MinLength', 25);

% 3. Highlight Transverse Fracture Line
figure; imshow(xray); hold on;
for k = 1:length(lines)
    xy = [lines(k).point1; lines(k).point2];
    plot(xy(:,1), xy(:,2), 'LineWidth', 2, 'Color', 'green');
end
title('Detected Bone Fracture Lines & Cortical Discontinuity');

% 1. Load Pre-Trained YOLOv4 Deep Endoscopy Detector
detector = load('trained_wce_polyp_yolov4.mat').detector;

% 2. Detect Polyps in Streaming Endoscopy Frame
frame = imread('capsule_endoscopy_frame104.jpg');
[bboxes, scores, labels] = detect(detector, frame, 'Threshold', 0.6);

% 3. Overlay Detection Bounding Boxes & Confidence
if ~isempty(bboxes)
    annotatedFrame = insertObjectAnnotation(frame, 'rectangle', bboxes, cellstr(labels));
    figure; imshow(annotatedFrame); title('Polyp Detected (High Confidence)');
end

% 1. Identify Metal Region via High Hounsfield Thresholding
ct_raw = dicomread('ct_hip_prosthesis.dcm');
metal_mask = ct_raw > 2500; % Metal HU threshold

% 2. Forward Radon Transform into Sinogram Space
theta = 0:179;
sino = radon(double(ct_raw), theta);
metal_sino = radon(double(metal_mask), theta);

% 3. Sinogram Inpainting & Filtered Backprojection (FBP)
sino_corrected = inpaintCoherent(sino, metal_sino > 0);
ct_recon = iradon(sino_corrected, theta);
ct_clean = imnlmfilt(mat2gray(ct_recon), 'DegreeOfSmoothing', 0.03);

% 1. Segment Lung Parenchyma using Optimal Threshold & Inversion
V_ct = dicomreadVolume('LIDC_CT_Series');
V_lung = (V_ct > -950) & (V_ct < -400); % Air/lung tissue window
V_lung = imfill(bwareaopen(V_lung, 5000), 'holes');

% 2. Extract 3D Geometric Regionprops on Nodule Candidates
cc = bwconncomp(V_lung);
stats3D = regionprops3(cc, 'Volume', 'EquivDiameter', 'PrincipalAxisLength');

% 3. Filter spheres with diameters between 3mm and 30mm
isNodule = stats3D.EquivDiameter >= 3.0 & stats3D.EquivDiameter <= 30.0;
noduleLocations = stats3D.Centroid(isNodule, :);

% 1. Fit Corneal Elevation to Polar Zernike Polynomials
[rho, theta] = meshgrid(linspace(0, 1, 200), linspace(0, 2*pi, 200));
[x, y] = pol2cart(theta, rho);

% 2. Compute Wavefront Aberration Surface
Z_coma = (3*rho.^3 - 2*rho) .* cos(theta); % 3rd order vertical coma
Z_spherical = 6*rho.^4 - 6*rho.^2 + 1;     % 4th order spherical
cornealElevation = 1.2*Z_coma + 0.8*Z_spherical;

figure; surfc(x*4, y*4, cornealElevation); shading interp; colormap jet;
title('3D Corneal Curvature Elevation Map (Diopters)');

% 1. Load Blood Film Micrograph and Transform to HSV Color Space
smear = imread('malaria_giemsa_smear.jpg');
hsv = rgb2hsv(smear);
sat = hsv(:,:,2); % Parasite chromatin stains strong violet/purple in saturation

% 2. Threshold Parasite Chromatin Spots
parasite_mask = sat > 0.45;
parasite_mask = bwareaopen(parasite_mask, 5); % Remove tiny camera noise

% 3. Segment Total Red Blood Cells
[centers, radii] = imfindcircles(rgb2gray(smear), [20 40]);
parasitemia = (sum(parasite_mask(:) > 0) / length(centers)) * 100;
fprintf('Total RBCs: %d | Parasitemia Level: %.2f%%\n', length(centers), parasitemia);