Deep Bayesian active learning-to-rank with relative annotation for estimation of ulcerative colitis severity

Takeaki Kadota, Hideaki Hayashi, Ryoma Bise, Kiyohito Tanaka, Seiichi Uchida

Abstract: Highlights•Propose a deep Bayesian active learning-to-rank with relative annotation.•Demonstrate the applicability of Monte Carlo dropout to a Siamese neural network.•Improve the estimation performance by selecting effective pairs for training.•Demonstrate that our method improves robustness to class imbalance.•Achieve high severity estimation performance in ranking and classification tasks.