Deep segmentation networks generally consist of an encoder to extract features from an input image and a decoder to restore them to the original input size to produce segmentation results. In an ideal setting, the trained encoder should possess the semantic embedding capability, which maps a pair of features close to each other when they belong to the same class, and maps them distantly if they correspond to different classes. Recent deep segmentation networks do not directly deal with the embedding behavior of the encoder. Accordingly, we cannot expect that the features embedded by the encoder will have the semantic embedding property. If the model can be trained to have the embedding ability, it will further enhance the performance as restoring from those features is much easier for the decoder. To this end, we propose supervised contrastive embedding, which employs feature-wise contrastive loss for the feature map to enhance the segmentation performance on medical images. We also introduce a boundary-aware sampling strategy, which focuses on the features corresponding to image patches located at the boundary area of the ground-truth annotations. Through extensive experiments on lung segmentation in chest radiographs, liver segmentation in computed tomography, and brain tumor and spinal cord gray matter segmentation in magnetic resonance images, it is demonstrated that the proposed method helps to improve the segmentation performance of popular U-Net, U-Net++, and DeepLabV3+ architectures. Furthermore, it is confirmed that the robustness on domain shifts can be enhanced for segmentation models by the proposed contrastive embedding.

Objective: The purpose of the current study was to investigate the diagnostic performance of Tc-99m methoxy-isobutyl-isonitrile (MIBI) for differentiation of malignant thyroid nodules (TN) through a systematic review and meta-analysis. Methods: The MEDLINE/PubMed and EMBASE database, from the earliest available date of indexing through January 31, 2018, were searched for studies evaluating the diagnostic performance of Tc-99m MIBI for TN. We determined the sensitivities and specificities across studies, calculated positive and negative likelihood ratios (LR+ and LR-), and constructed summary receiver operating characteristic curves. Results: Across 22 studies (2,421 patients), the pooled sensitivity for Tc-99m MIBI thyroid scan was 0.87 (95% CI; 0.76-0.93) with heterogeneity (I2=92.3) and a pooled specificity of 0.78 (95% CI; 0.67-0.86) with heterogeneity (I2=96.4). Likelihood ratio (LR) syntheses gave an overall positive likelihood ratio (LR+) of 4.0 (95% CI; 2.5-6.3) and negative likelihood ratio (LR-) of 0.17 (95% CI; 0.09-0.32). The pooled diagnostic odds ratio (DOR) was 24 (95% CI; 63-176). Hierarchical summary receiver operating characteristic (ROC) curve and indicates that the areas under the curve was 0.90 (95% CI; 0.87-0.92). In meta-regression analysis, no definite variable was the source of the study heterogeneity. Conclusion: The current meta-analysis showed the moderate sensitivity and specificity of Tc-99m MIBI thyroid scan for differentiation of malignant TN. However, the DOR was relatively low and the likelihood ratio scatter-gram indicated that Tc-99m MIBI thyroid scan is not useful for confirmation of presence of malignant TN and not for its exclusion. Therefore, cautious application and interpretation should be paid to the Tc-99m MIBI thyroid scan when differentiating malignant TN.

Fine-grained image recognition aims to classify fine subcategories belonging to the same parent category, such as vehicle model or bird species classification. This is an inherently challenging task because a classifier must capture subtle interclass differences under large intraclass variances. Most previous approaches are based on supervised learning, which requires a large-scale labeled dataset. However, such large-scale annotated datasets for fine-grained image recognition are difficult to collect because they generally require domain expertise during the labeling process. In this study, we propose a self-supervised transfer learning method based on Vision Transformer (ViT) to learn finer representations without human annotations. Interestingly, it is observed that existing self-supervised learning methods using ViT (e.g., DINO) show poor patch-level semantic consistency, which may be detrimental to learning finer representations. Motivated by this observation, we propose a consistency loss function that encourages patch embeddings of the overlapping area between two augmented views to be similar to each other during self-supervised learning on fine-grained datasets. In addition, we explore effective transfer learning strategies to fully leverage existing self-supervised models trained on large-scale labeled datasets. Contrary to the previous literature, our findings indicate that training only the last block of ViT is effective for self-supervised transfer learning. We demonstrate the effectiveness of our proposed approach through extensive experiments using six fine-grained image classification benchmark datasets, including FGVC Aircraft, CUB-200-2011, Food-101, Oxford 102 Flowers, Stanford Cars, and Stanford Dogs. Under the linear evaluation protocol, our method achieves an average accuracy of $78.5\%$, outperforming the existing transfer learning method, which yields $77.2\%$.