[TMI2019-02]Learning Cross-Modality Representations From Multi-Modal Images