In The American journal of pathology ; h5-index 54.0
Osteosarcomas are aggressive bone tumors with many divergent histologic patterns. During pathology review, osteosarcomas are subtyped based on the predominant histologic pattern; however, tumors often demonstrate multiple patterns. This high tumor heterogeneity coupled with scarcity of samples compared to other tumor types render histology-based prognosis of osteosarcomas challenging. To combat lower-case numbers in humans, dogs with spontaneous osteosarcomas have been suggested as a model species. Here, we adversarially train a convolutional neural network to classify distinct histological patterns of osteosarcoma in humans using mostly canine Osteosarcoma data during training. We show that adversarial training improves domain adaption of a histologic subtype classifier from canines to humans achieving an average multi-class F1 score of 0.77 (CI: 0.74-0.79) and 0.80 (CI: 0.78-0.81) when compared to the ground truth in canines and humans, respectively. Finally, we applied our trained model to characterize the histologic landscape of 306 canine osteosarcomas and uncovered distinct clusters with markedly different clinical responses to standard of care therapy.
Patkar Sushant, Beck Jessica, Harmon Stephanie, Mazcko Christina, Turkbey Baris, Choyke Peter, Brown G Tom, LeBlanc Amy
2022-Oct-26
