| 主权项 |
1. An unsupervised computer-implemented method of segmenting a digital image of biological tissue, the computer including a processor, the method comprising acts of:
generating, by the processor, a first weak unsupervised data driven segmentation of the digital image of the biological tissue to segment the digital image into a first set of biological units where the biological units are the boundary of each cell of the biological tissue; assigning, by the processor, an unsupervised individual rank to each biological unit in the first set of biological units based on a similarity in shape and scale between the biological unit and one or more other biological units in the first set of biological units, wherein a geometric model is used to define the similarity as a function of a weighted sum of the similarity in shape and scale between each biological unit and the one or more other biological units; and selecting, by the processor, a subset of biological units from the first set of biological units, where the subset of biological units are boundaries of cells in the biological tissue, based on the unsupervised individual rank of each biological unit relative to a predetermined threshold rank; generating, by the processor, a second weak unsupervised data driven segmentation of only a portion of the digital image including the subset of biological units to segment the portion of the digital image into a fully unsupervised second set of biological units; and repeating the acts of assigning the rank, selecting the subset and generating the second weak unsupervised data driven segmentation until an average change in the rank of substantially all of the biological units in the digital image is no greater than a predetermined threshold level subsequent to the act of generating the second weak unsupervised data driven segmentation, wherein the second weak or data driven segmentation is different each time the second weak or data driven segmentation is generated. |
