Distribution Matching Losses Can Hallucinate Features in Medical Image Translation
We discuss how image translation research can lead to the misdiagnosis of medical conditions. This can cause an issue when the data provided in the target domain has an over or under representation of some classes (e.g. healthy or sick). We recommend that these translated images should not be used for direct interpretation (e.g. by doctors).
Joseph Paul Cohen, Margaux Luck, Sina Honari. “Distribution Matching Losses Can Hallucinate Features in Medical Image Translation.” Medical Image Computing & Computer Assisted Intervention (MICCAI), 2018.