FIG. 3C further illustrates some embodiments discussed above. FIG. 3C shows superpixel generator 110 obtaining an exemplary biological sample image 310 and producing a plurality of superpixels 320 based on biological sample image 310. Feature extractor 111 can then obtain the plurality of superpixels 320 and extract feature vectors 330 for all the superpixels. Clustering engine 112 can obtain feature vectors 330, and based at least on feature vectors 330, generated a predefined number of (in this example, four) clusters A, B, C, and D arranged, for example, as illustrated in clustering information 350 and visualized by clustering information image 340. As discussed above, each cluster can be characterized or represented by a centroid vector, as illustrated in exemplary centroid vector table 370. Clustering engine 112 can also precalculate distances between each pair of different clusters, as illustrated in exemplary distance table 360. Storage interface 113 can then store clustering information 350 into memory 116 (or any other memory). As discussed above, storage interface 113 can also store into memory distance table 360 and/or centroid vector table 370.

FIG. 3D shows graph engine 112 retrieving from memory 116 (through storage interface 113) clustering information 350, and distance table 360 and/or centroid vector table 370. Based at least on this information, graph engine 112 generates an exemplary graph 380.