New feature : vessels mask

[TitouanGragnic]

Here is a detailed step by step guide on how to mplement the vessels mask (vein + artery):

I/ First Masks and Correlation

• [x] Apply a flat field correction on the moment 0
• [ ] Get a diaphragm mask sqrt((X - numX / 2) .^ 2 + (Y - numY / 2) .^ 2) <= 0.45 * (numY + numX) / 2;

1) Compute vesselness response

• [x] Get rid of singletons from mean of the moment 0 "flat fielded" mean(M0_disp_video, 3)
• [ ] Filter vesselness (from the above result) then use it to get a first vesselness mask by using Otsu algorithm

2) Compute first correlation

Check createMasks.m in PulseWave line 38

• [ ] Compute pulse vascularPulse = sum(M0_disp_video .* maskVesselness, [1 2]); vascularPulse = vascularPulse ./ nnz(maskVesselness); vascularPulse_centered = vascularPulse - mean(vascularPulse, 3);
• [ ] Get the Correlation Matrix of VascularPulse (mean on standart variation) mean(M0_disp_centered_video .* vascularPulse_centered, 3) ./ (std((M0_disp_centered_video), [], 3) * std(vascularPulse_centered, [], 3));

3) Compute the barycentres and the circle mask

Physisict said we already have the varycentres, need more specification from them

4) Segment Vessels

• [ ] Fine tune the previous vesselness mask by using bwareafilt wich extract objects from binary image by size.
• [ ] Build classes and their Threshold, no more specification

Check createMasks.m line 60 for more details

5) Segment Arteries and Veins

• [ ] Get first version of vein and artery mask from the previous classes firstMaskArtery = quantizedVesselCorrelation == 4 |quantizedVesselCorrelation == 5; firstMaskVein = quantizedVesselCorrelation == 3 | quantizedVesselCorrelation == 2;

• [ ] Use bwareaopen to remove small 'blobs' from the two masks

II/ Compute the second correlation to fine tune the segmentation

1) Compute the new pulse

• [ ] Same as I/ 2) but for arterialPulse and venousPulse to get correlation matrix of Arterial pulse and venous pulse

Let's do that and we'll see after for the cleaning

Need to do interpolation: 2 suggestions

• NPPI (probably too slow)
• Zero padding FFT

[TitouanGragnic]

Needed function implemented (in order):

I didn't incuded all the functions of the Matlap code because I'm not sure if they are all relevant, I only inculded those I tought were obvious

• [x] Flat field correction @alexispinson and @GustaveHerve
• [ ] Vesselness filter of Frangi @NaN
• [x] Otsu algorithm @TitouanGragnic
• [ ] bwareafilt @TitouanGragnic
• [ ] bwareaopen @NaN

[TitouanGragnic]

otsu is working in branch add_ostsu