C2M-DoT: Cross-modal consistent multi-view medical report generation with domain transfer network

Ruizhi Wang, Zhenghua Xu, Xiangtao Wang, Weipeng Liu, Thomas Lukasiewicz

Abstract: Highlights•Propose a novel cross-modal model (C2M-DoT) to better generate medical reports.•Propose a multi-view contrastive learning strategy to utilize multi-view information.•Propose a domain transfer network to get good performance using single-view inputs.•Propose a cross-modal optimization (CMC) loss to better learn visual semantics.•Extensive experiments prove the effectiveness of C2M-DoT upon the existing baselines.

External IDs:dblp:journals/inffus/WangXWLL26