Diffusion-based deep learning method for augmenting ultrastructural imaging and volume electron microscopy.

Electron microscopy (EM) revolutionized the way to visualize cellular ultrastructure. Volume EM (vEM) has further broadened its three-dimensional nanoscale imaging capacity. However, intrinsic trade-offs between imaging speed and quality of EM restrict the attainable imaging area and volume. Isotropic imaging with vEM for large biological volumes remains unachievable. Here, we developed EMDiffuse, a suite of algorithms designed to enhance EM and vEM capabilities, leveraging the cutting-edge image generation diffusion model. EMDiffuse generates realistic predictions with high resolution ultrastructural details and exhibits robust transferability by taking only one pair of images of 3 megapixels to fine-tune in denoising and super-resolution tasks. EMDiffuse also demonstrated proficiency in the isotropic vEM reconstruction task, generating isotropic volume even in the absence of isotropic training data. We demonstrated the robustness of EMDiffuse by generating isotropic volumes from seven public datasets obtained from different vEM techniques and instruments. The generated isotropic volume enables accurate three-dimensional nanoscale ultrastructure analysis. EMDiffuse also features self-assessment functionalities on predictions' reliability. We envision EMDiffuse to pave the way for investigations of the intricate subcellular nanoscale ultrastructure within large volumes of biological systems.

A Unified Visual Information Preservation Framework for Self-supervised Pre-Training in Medical Image Analysis.

Recent advances in self-supervised learning (SSL) in computer vision are primarily comparative, whose goal is to preserve invariant and discriminative semantics in latent representations by comparing siamese image views. However, the preserved high-level semantics do not contain enough local information, which is vital in medical image analysis (e.g., image-based diagnosis and tumor segmentation). To mitigate the locality problem of comparative SSL, we propose to incorporate the task of pixel restoration for explicitly encoding more pixel-level information into high-level semantics. We also address the preservation of scale information, a powerful tool in aiding image understanding but has not drawn much attention in SSL. The resulting framework can be formulated as a multi-task optimization problem on the feature pyramid. Specifically, we conduct multi-scale pixel restoration and siamese feature comparison in the pyramid. In addition, we propose non-skip U-Net to build the feature pyramid and develop sub-crop to replace multi-crop in 3D medical imaging. The proposed unified SSL framework (PCRLv2) surpasses its self-supervised counterparts on various tasks, including brain tumor segmentation (BraTS 2018), chest pathology identification (ChestX-ray, CheXpert), pulmonary nodule detection (LUNA), and abdominal organ segmentation (LiTS), sometimes outperforming them by large margins with limited annotations. Codes and models are available at https://github.com/RL4M/PCRLv2.

GraVIS: Grouping Augmented Views From Independent Sources for Dermatology Analysis.

Self-supervised representation learning has been extremely successful in medical image analysis, as it requires no human annotations to provide transferable representations for downstream tasks. Recent self-supervised learning methods are dominated by noise-contrastive estimation (NCE, also known as contrastive learning), which aims to learn invariant visual representations by contrasting one homogeneous image pair with a large number of heterogeneous image pairs in each training step. Nonetheless, NCE-based approaches still suffer from one major problem that is one homogeneous pair is not enough to extract robust and invariant semantic information. Inspired by the archetypical triplet loss, we propose GraVIS, which is specifically optimized for learning self-supervised features from dermatology images, to group homogeneous dermatology images while separating heterogeneous ones. In addition, a hardness-aware attention is introduced and incorporated to address the importance of homogeneous image views with similar appearance instead of those dissimilar homogeneous ones. GraVIS significantly outperforms its transfer learning and self-supervised learning counterparts in both lesion segmentation and disease classification tasks, sometimes by 5 percents under extremely limited supervision. More importantly, when equipped with the pre-trained weights provided by GraVIS, a single model could achieve better results than winners that heavily rely on ensemble strategies in the well-known ISIC 2017 challenge. Code is available at https://bit.ly/3xiFyjx.