Abstract:
Pathology has become increasingly more reliant on digital imaging as a
means for viewing, sharing, and archiving slides, and as an essential
first step for the application of advanced image analysis to support
cancer diagnostics. In H&E stained tissue, cell nuclei are especially
prominent, and their shapes, staining attributes, and distributions
within the tissue serve as important diagnostic and prognostic features.
Therefore, the ability to accurately identify and segment nuclei from
other tissue structures is paramount toward developing a reliable
analytical tool. We developed an algorithm that rapidly identifies
candidate nuclei and segments them in a manner that retains much of the
shape information and location precision. The algorithm uses color
analysis, template matching based on shape, and clump splitting to
demarcate individual nuclei and to segregate overlapping nuclei. Given
its speed and relative simplicity, this method is especially amenable
to processing large image regions at high magni cation, making high
throughput and on-demand analysis realizable.
