This code merges the first two dimensions of a higher dimensional array (e.g. an image stack) into a single dimension. The difference to a simple reshape is that the output matrix only contains selected pixels in a region of interest (ROI) by using a 2D logical index (e.g. from thresholding a single image). The reduced matrix only contains the values of interest to increase analysis speed and reduce data size. Lastly, the original data structure can be recovered after analysis. When reconstructing, all values that were not part of the initial ROI are recreated as NaNs.