Method for determining the short axis in a lesion region in a three dimensional medical image

A computer implemented method for identifying a short axis in an image of a lesion region, the method comprising the steps of:

• inputting a set of voxels defining a 3D lesion, and defining a long axis of the lesion and a direction of a plane including the long axis and a short axis;
• dividing the long axis into a plurality of ranges, each of the plurality of ranges including a preset range width;
• projecting each point of the input set of voxels onto the long axis in a direction perpendicular or substantially perpendicular to a direction of the long axis;
• forming clusters of points projected onto a same range and, within each of the formed clusters, defining a most distant point pair;
• selecting the most distant pair among the formed clusters as intra_candidate;
• for each of the plurality of ranges, finding inter_candidates by evaluating pairs of points with one point belonging to a first range and another point belonging to an adjacent range and selecting the most distant pair of the pairs of points for each of the plurality of ranges;
• selecting among the inter_candidates a most distant pair pair_inter satisfying a condition that through its voxels an axis is able to be drawn that is perpendicular or substantially perpendicular to the long axis with an angle of deviation from the direction perpendicular or substantially perpendicular to the long axis that is smaller than a pre-defined value;
• assigning the short axis either to the intra_candidate or to the pair_inter if the short axis through its points is longer than an axis assigned to the inter_candidate;
• and
• if no pair is found for which inside the voxels of the pair an axis is able to be drawn that is perpendicular or substantially perpendicular to the long axis, increasing the preset range width and repeating the method from the dividing step.