Binary mask for .stl file of large size

I'm trying to create binary mask for .stl files of segemented organs. With fat and muscle data of size (3116232x3 double) and (1521380x3 double) respectively, the inpolyhedron function doesn't work and shows an error saying it goes beyond 15 GB. So, I try to reduce their size. Which works fine, but on doing so a large amount of data is cropped out. Your suggestions would be valuable.

vertices = triangulationData.Points;
faces = triangulationData.ConnectivityList;
      
if size(vertices, 1) > 1e6 || size(faces, 1) > 1e6
    [reducedFaces, reducedVertices] = reducepatch(struct('faces', faces, 'vertices', vertices), 5);
    points = reducedVertices;
    faces = reducedFaces;
else
    points = vertices;
    faces = faces;
end

Answer 1

[Ronit]

Hello Toshini,

To address the issue of creating binary masks for large ".stl" files, I recommend using a chunk-based approach. This method allows processing large datasets in smaller sections, reducing memory usage and avoiding significant data loss from mesh reduction. This approach maintains the original detail and quality of the mesh, ensuring that important features are not lost.

Here is an example on how to implement a chunk-based approach:

vertices = triangulationData.Points;
faces = triangulationData.ConnectivityList;

chunkSize = 1e6;  % Adjust based on available memory
binaryMask = zeros(size(vertices, 1), 1);

% Process in chunks
numChunks = ceil(size(vertices, 1) / chunkSize);

for i = 1:numChunks
    % Determine the range for the current chunk and extract the vertices
    startIdx = (i - 1) * chunkSize + 1;
    endIdx = min(i * chunkSize, size(vertices, 1));
    chunkVertices = vertices(startIdx:endIdx, :);
    
    insideMask = inpolyhedron(faces, vertices, chunkVertices);
    binaryMask(startIdx:endIdx) = insideMask;
end

Note: Adjust the above code according to the dataset and specific requirements of your application, as different datasets may require modifications to chunk size, memory allocation, or processing logic.

I hope this helps you!