Unsupervised shape-and-texture-based generative adversarial tuning of pre-trained networks for carotid segmentation from 3D ultrasound images.

BACKGROUND: Vessel-wall volume and localized three-dimensional ultrasound (3DUS) metrics are sensitive to the change of carotid atherosclerosis in response to medical/dietary interventions. Manual segmentation of the media-adventitia boundary (MAB) and lumen-intima boundary (LIB) required to obtain these metrics is time-consuming and prone to observer variability. Although supervised deep-learning segmentation models have been proposed, training of these models requires a sizeable manually segmented training set, making larger clinical studies prohibitive. PURPOSE: We aim to develop a method to optimize pre-trained segmentation models without requiring manual segmentation to supervise the fine-tuning process. METHODS: We developed an adversarial framework called the unsupervised shape-and-texture generative adversarial network (USTGAN) to fine-tune a convolutional neural network (CNN) pre-trained on a source dataset for accurate segmentation of a target dataset. The network integrates a novel texture-based discriminator with a shape-based discriminator, which together provide feedback for the CNN to segment the target images in a similar way as the source images. The texture-based discriminator increases the accuracy of the CNN in locating the artery, thereby lowering the number of failed segmentations. Failed segmentation was further reduced by a self-checking mechanism to flag longitudinal discontinuity of the artery and by self-correction strategies involving surface interpolation followed by a case-specific tuning of the CNN. The U-Net was pre-trained by the source dataset involving 224 3DUS volumes with 136, 44, and 44 volumes in the training, validation and testing sets. The training of USTGAN involved the same training group of 136 volumes in the source dataset and 533 volumes in the target dataset. No segmented boundaries for the target cohort were available for training USTGAN. The validation and testing of USTGAN involved 118 and 104 volumes from the target cohort, respectively. The segmentation accuracy was quantified by Dice Similarity Coefficient (DSC), and incorrect localization rate (ILR). Tukey's Honestly Significant Difference multiple comparison test was employed to quantify the difference of DSCs between models and settings, where p ≤ 0.05 $p\,\le \,0.05$ was considered statistically significant. RESULTS: USTGAN attained a DSC of 85.7 ± 13.0 $85.7\,\pm \,13.0$ % in LIB and 86.2 ± 10.6 ${86.2}\,\pm \,{10.6}$ % in MAB, improving from the baseline performance of 74.6 ± 30.7 ${74.6}\,\pm \,{30.7}$ % in LIB (p < 10 - 12 $<10^{-12}$ ) and 75.7 ± 28.9 ${75.7}\,\pm \,{28.9}$ % in MAB (p < 10 - 12 $<10^{-12}$ ). Our approach outperformed six state-of-the-art domain-adaptation models (MAB: p ≤ 3.63 × 10 - 7 $p \le 3.63\,\times \,10^{-7}$ , LIB: p ≤ 9.34 × 10 - 8 $p\,\le \,9.34\,\times \,10^{-8}$ ). The proposed USTGAN also had the lowest ILR among the methods compared (LIB: 2.5%, MAB: 1.7%). CONCLUSION: Our framework improves segmentation generalizability, thereby facilitating efficient carotid disease monitoring in multicenter trials and in clinics with less expertise in 3DUS imaging.

Global Analysis of 2-Hydroxy Fatty Acids by Gas Chromatography-Tandem Mass Spectrometry Reveals Species-Specific Enrichment in Echinoderms.

Abnormal levels of 2-hydroxy fatty acids (2-OH FAs) are characterized in multiple diseases, and their quantification in foodstuffs is critical to identify the sources of supplementation for potential treatment. However, due to the structural complexity and limited available standards, the comprehensive profiling of 2-OH FAs remains an ongoing challenge. Herein, an innovative approach based on gas chromatography-tandem mass spectrometry (GC-MS/MS) was developed to determine the full profile of these FA metabolites. MS and MS/MS spectra of the trimethylsilyl (TMS) derivatives of 2-OH fatty acid methyl esters (FAMEs) were collected for peak annotation by their signature fragmentation patterns. The structures were further confirmed by validated structure-dependent retention time (RT) prediction models, taking advantage of the correlation between the RT, carbon chain length, and double bond number from commercial standards and pseudostandards identified in the whole-brain samples from mice. An in-house database containing 50 saturated and monounsaturated 2-OH FAs was established, which is expandible when additional molecular species with different chain lengths and backbone structures are identified in the future. A quantitation method was then developed by scheduled multiple reaction monitoring (MRM) and applied to investigate the profiling of 2-OH FAs in echinoderms. Our results revealed that the levels of total 2-OH FAs in sea cucumber Apostichopus japonicas (8.40 ± 0.28 mg/g dry weight) and starfish Asterias amurensis (7.51 ± 0.18 mg/g dry weight) are much higher than that in sea urchin Mesocentrotus nudus (531 ± 108 µg/g dry weight). Moreover, 2-OH C24:1 is the predominant molecular species accounting for 67.9% of the total 2-OH FA in sea cucumber, while 2-OH C16:0 is the major molecular species in starfish. In conclusion, the current innovative GC-MS approach has successfully characterized distinct molecular species of 2-OH FAs that are highly present in sea cucumbers and starfish. Thus, these findings suggest the possibility of developing future feeding strategies for preventing and treating diseases associated with 2-OH FA deficiency.

Meta-Transfer Learning Through Hard Tasks.

Meta-learning has been proposed as a framework to address the challenging few-shot learning setting. The key idea is to leverage a large number of similar few-shot tasks in order to learn how to adapt a base-learner to a new task for which only a few labeled samples are available. As deep neural networks (DNNs) tend to overfit using a few samples only, typical meta-learning models use shallow neural networks, thus limiting its effectiveness. In order to achieve top performance, some recent works tried to use the DNNs pre-trained on large-scale datasets but mostly in straight-forward manners, e.g., (1) taking their weights as a warm start of meta-training, and (2) freezing their convolutional layers as the feature extractor of base-learners. In this paper, we propose a novel approach called meta-transfer learning (MTL), which learns to transfer the weights of a deep NN for few-shot learning tasks. Specifically, meta refers to training multiple tasks, and transfer is achieved by learning scaling and shifting functions of DNN weights (and biases) for each task. To further boost the learning efficiency of MTL, we introduce the hard task (HT) meta-batch scheme as an effective learning curriculum of few-shot classification tasks. We conduct experiments for five-class few-shot classification tasks on three challenging benchmarks, miniImageNet, tieredImageNet, and Fewshot-CIFAR100 (FC100), in both supervised and semi-supervised settings. Extensive comparisons to related works validate that our MTL approach trained with the proposed HT meta-batch scheme achieves top performance. An ablation study also shows that both components contribute to fast convergence and high accuracy.