Robert-von-Ostertag-Str. 15
14163 Berlin
+49 30 838 62450
pathologie@vetmed.fu-berlin.de
Introduction:
Anisokaryosis is a prognostic criterion for many tumor types. Although it is traditionally estimated by pathologists, measurements have advantages for statistical evaluation and may improve reproducibility. The aim of the study was to compare these two methods in canine cutaneous mast cell tumors (ccMCT).
Material and Methods:
A Deep learning-based algorithm was utilized to calculate the standard deviation (SD) of nuclear size in histologic images of 96 ccMCT with known outcome. Three pathologists estimated the degree (low, moderate, high) of anisokaryosis in the same images.
Results:
The algorithm predicted tumor-specific survival with a sensitivity of 85% and specificity of 89% and a hazard ratio (HR) of 26.7 (p < 0.001) at a cut-off of SD = 10.15 µm². All three pathologists estimated the same anisokaryosis category in 35% of the cases. High anisokaryosis had a sensitivity and specificity of 38% and 86%, 46% and 87%, 62% and 96% for the individual pathologists, respectively. The HR ranged between 3.0–21.5.
Conclusion:
We have shown a strong prognostic value of pathologist estimates and algorithmic measurements of anisokaryosis in these ccMCT. Measurements of anisokaryosis may be more advantageous given the inherent ability to balance sensitivity and specificity.