2.75D: Boosting learning by representing 3D Medical imaging to 2D features for small data
This work addresses the problem of efficient and effective 3D medical image analysis for researchers and practitioners by offering a novel representation that reduces computational demands and leverages existing 2D models, though it is incremental in improving upon existing 2D and 3D methods.
The authors tackled the challenge of applying 3D CNNs in medical imaging due to high computational costs and data scarcity by proposing 2.75D, a method that represents 3D volumetric data as 2D features using a spiral-spinning technique, enabling the use of pre-trained 2D CNNs and achieving significant performance gains, such as outperforming 2D, 2.5D, and 3D counterparts on a lung dataset and reducing training and inference time substantially.
In medical-data driven learning, 3D convolutional neural networks (CNNs) have started to show superior performance to 2D CNNs in numerous deep learning tasks, proving the added value of 3D spatial information in feature representation. However, the difficulty in collecting more training samples to converge, more computational resources and longer execution time make this approach less applied. Also, applying transfer learning on 3D CNN is challenging due to a lack of publicly available pre-trained 3D models. To tackle these issues, we proposed a novel 2D strategical representation of volumetric data, namely 2.75D. In this work, the spatial information of 3D images is captured in a single 2D view by a spiral-spinning technique. As a result, 2D CNN networks can also be used to learn volumetric information. Besides, we can fully leverage pre-trained 2D CNNs for downstream vision problems. We also explore a multi-view 2.75D strategy, 2.75D 3 channels (2.75Dx3), to boost the advantage of 2.75D. We evaluated the proposed methods on three public datasets with different modalities or organs (Lung CT, Breast MRI, and Prostate MRI), against their 2D, 2.5D, and 3D counterparts in classification tasks. Results show that the proposed methods significantly outperform other counterparts when all methods were trained from scratch on the lung dataset. Such performance gain is more pronounced with transfer learning or in the case of limited training data. Our methods also achieved comparable performance on other datasets. In addition, our methods achieved a substantial reduction in time consumption of training and inference compared with the 2.5D or 3D method.