2.75D: Boosting learning by representing 3D Medical imaging to 2D features for small data

Xin Wang, Ruisheng Su, Weiyi Xie, Wenjin Wang, Yi Xu, Ritse Mann, Jungong Han, Tao Tan

Abstract: Highlights•We propose a 2.75D strategy, which uses the spiral scanning technique to extract efficient 2D representations of a 3D volume.•We evaluate the proposed methods on three public datasets with different modalities and different organs and demonstrate its remarkable performance in comparison to 2D, 2.5D, and 3D approaches.•We explore the capability and advantage of 2.75D using transfer learning.•We systematically investigate the effect of data size on training yield from different approaches.