Extended Data Fig. 3: Flowchart for morphology and displacement-based global cost minimization tracking.

a, A matrix of intensity weighted centroid distances is built comprised of the difference between every existing track (columns) and every mitochondrion found in the current time frame (rows). A region of mitochondria from an adjacent cell can be seen in the dotted white area, resulting in large differences between the mitochondria of the two cells. b, A mask is created to avoid assignments between any mitochondria and track which are further than 3 frames apart (the maximum search time threshold), and at a distance greater than 1 μm multiplied by the frame time apart (the maximum velocity threshold). c, The mask is applied to the difference matrices of all mitochondrial features and distance. d, The difference matrices are then squared, z score normalized, weighted, and summed to create the final difference matrix (e). f, A diagonal matrix of new track costs is appended to the difference matrix to allow newly found mitochondria to be assigned to a new track. g, Every mitochondrion (row) in the matrix is assigned to a single track (column) that leads to a globally minimized cost of assignment. h, A mitochondrion which is assigned to a track that also corresponds to its own individually minimum cost is labelled as a confident assignment. i, This process is repeated for all temporal frames. All confident tracks are then analyzed for interCoV:intraCoV ratios for each mitochondrial morphological feature, and weights are set accordingly. These weights are used to rerun the tracking algorithm, leading to a higher number of mitochondria that are confidently tracked for the full length of the time stack (j) and to less total number of tracks overall (k).