Joint categorical and ordinal learning for cancer grading in pathology images

• Published in: Medical image analysis 2021-10, Vol.73, p.102206-102206, Article 102206
• Main Authors: Vuong, Trinh Thi Le, Kim, Kyungeun, Song, Boram, Kwak, Jin Tae
• Format: Article
• Language: English
• Subjects: Cancer; Cancer grading; Categorical classification; Classification; Image acquisition; Image analysis; Image classification; Image processing; Learning; Multi-task learning; Ordinal classification; Pathology; Prostate
• ISSN: 1361-8415; 1361-8423
• DOI: 10.1016/j.media.2021.102206

•Propose a deep neural network for cancer grading that takes advantages of both categorical classification and ordinal classification.•Introduce a new loss function for the ordinal classification problem that could aid in improving the accuracy and generalizability of the network.•Colorectal and prostate cancer grading are successfully conducted and outperform 9 other competing models.•Tissue image datasets that underwent different acquisition and processing procedures are employed to assess the generalizability of the proposed network. [Display omitted] Cancer grading in pathology image analysis is one of the most critical tasks since it is related to patient outcomes and treatment planning. Traditionally, it has been considered a categorical problem, ignoring the natural ordering among the cancer grades, i.e., the higher the grade is, the more aggressive it is, and the worse the outcome is. Herein, we propose a joint categorical and ordinal learning framework for cancer grading in pathology images. The approach simultaneously performs both categorical classification and ordinal classification and aims to leverage the distinctive features from the two tasks. Moreover, we propose a new loss function for the ordinal classification task that offers an improved contrast between the correctly classified examples and misclassified examples. The proposed method is evaluated on multiple collections of colorectal and prostate pathology images that underwent different acquisition and processing procedures. Both quantitative and qualitative assessments of the experimental results confirm the effectiveness and robustness of the proposed method in comparison to other competing methods. The results suggest that the proposed approach could permit improved histopathologic analysis of cancer grades in pathology images.