TransUNet: Rethinking the U-Net architecture design for medical image segmentation through the lens of transformers.

Medical image segmentation is crucial for healthcare, yet convolution-based methods like U-Net face limitations in modeling long-range dependencies. To address this, Transformers designed for sequence-to-sequence predictions have been integrated into medical image segmentation. However, a comprehensive understanding of Transformers' self-attention in U-Net components is lacking. TransUNet, first introduced in 2021, is widely recognized as one of the first models to integrate Transformer into medical image analysis. In this study, we present the versatile framework of TransUNet that encapsulates Transformers' self-attention into two key modules: (1) a Transformer encoder tokenizing image patches from a convolution neural network (CNN) feature map, facilitating global context extraction, and (2) a Transformer decoder refining candidate regions through cross-attention between proposals and U-Net features. These modules can be flexibly inserted into the U-Net backbone, resulting in three configurations: Encoder-only, Decoder-only, and Encoder+Decoder. TransUNet provides a library encompassing both 2D and 3D implementations, enabling users to easily tailor the chosen architecture. Our findings highlight the encoder's efficacy in modeling interactions among multiple abdominal organs and the decoder's strength in handling small targets like tumors. It excels in diverse medical applications, such as multi-organ segmentation, pancreatic tumor segmentation, and hepatic vessel segmentation. Notably, our TransUNet achieves a significant average Dice improvement of 1.06% and 4.30% for multi-organ segmentation and pancreatic tumor segmentation, respectively, when compared to the highly competitive nn-UNet, and surpasses the top-1 solution in the BrasTS2021 challenge. 2D/3D Code and models are available at https://github.com/Beckschen/TransUNet and https://github.com/Beckschen/TransUNet-3D, respectively.

Consistency-guided Differential Decoding for Enhancing Semi-supervised Medical Image Segmentation.

Semi-supervised learning (SSL) has been proven beneficial for mitigating the issue of limited labeled data, especially on volumetric medical image segmentation. Unlike previous SSL methods which focus on exploring highly confident pseudo-labels or developing consistency regularization schemes, our empirical findings suggest that differential decoder features emerge naturally when two decoders strive to generate consistent predictions. Based on the observation, we first analyze the treasure of discrepancy in learning towards consistency, under both pseudo-labeling and consistency regularization settings, and subsequently propose a novel SSL method called LeFeD, which learns the feature-level discrepancies obtained from two decoders, by feeding such information as feedback signals to the encoder. The core design of LeFeD is to enlarge the discrepancies by training differential decoders, and then learn from the differential features iteratively. We evaluate LeFeD against eight state-of-the-art (SOTA) methods on three public datasets. Experiments show LeFeD surpasses competitors without any bells and whistles, such as uncertainty estimation and strong constraints, as well as setting a new state of the art for semi-supervised medical image segmentation. Code has been released at https://github.com/maxwell0027/LeFeD.

UniMiSS+: Universal Medical Self-Supervised Learning From Cross-Dimensional Unpaired Data.

Self-supervised learning (SSL) opens up huge opportunities for medical image analysis that is well known for its lack of annotations. However, aggregating massive (unlabeled) 3D medical images like computerized tomography (CT) remains challenging due to its high imaging cost and privacy restrictions. In our pilot study, we advocated bringing a wealth of 2D images like chest X-rays as compensation for the lack of 3D data, aiming to build a universal medical self-supervised representation learning framework, called UniMiSS. Especially, we designed a pyramid U- like medical Transformer (MiT) as the backbone to make UniMiSS possible to perform SSL with both 2D and 3D images. Consequently, the predecessor UniMiSS has two obvious merits compared to current 3D-specific SSL: (1) more effective - superior to learning strong representations, benefiting from more and diverse data; and (2) more versatile - suitable for various downstream tasks without the restriction on the dimensionality barrier. Unfortunately, UniMiSS did not dig deeply into the intrinsic anatomy correlation between 2D medical images and 3D volumes due to the lack of paired multi-modal/dimension patient data. In this extension paper, we propose the UniMiSS+, in which we introduce the digitally reconstructed radiographs (DRR) technology to simulate X-ray images from a CT volume to access paired CT and X-ray data. Benefiting from the paired group, we introduce an extra pair- wise constraint to boost the cross-modality correlation learning, which also can be adopted as a cross-dimension regularization to further improve the representations. We conduct expensive experiments on multiple 3D/2D medical image analysis tasks, including segmentation and classification. The results show that the proposed UniMiSS+ achieves promising performance on various downstream tasks, not only outperforming the ImageNet pre-training and other advanced SSL counterparts substantially but also improving the predecessor UniMiSS pre-training. Code is available at: https://github.com/YtongXie/UniMiSS-code.

Formulation and characterization of nanocapsules loaded with roselle anthocyanins extract and enhancement of anthocyanins bioaccessibility.

Hibiscus sabdariffa L. (roselle) is a medicinal and edible plant which rich in anthocyanins with potent antioxidant properties. To enhance the stability of roselle anthocyanins, they were encapsulated in nanocapsules composed of carboxymethyl chitosan (CMC), chitosan hydrochloride (CHC), and ß-lactoglobulin (ß-Lg). In vitro simulated digestion assays evaluated the impact of various core-to-wall ratios and ß-Lg concentrations on the bioaccessibility of seven anthocyanins. Nanocapsules with a core-to-wall ratio of 1:2 and ß-Lg at 10 mg/mL exhibited the highest encapsulation efficiency (EE). Cyanidin-3-glucoside had the highest EE, while cyanidin-3-sambubioside showed the outstanding retention rate. Furthermore, simulated digestion experiments combined with molecular docking revealed that peonidin-3-glucoside and petunidin-3-glucoside likely interact with and bind to the outer ß-Lg layer of the nanocapsules, increasing their release during in vitro digestion. This study demonstrates that encapsulating roselle anthocyanins in CMC, CHC, and ß-Lg nanocapsules significantly enhances their bioaccessibility.

Factors and Perceptions Associated with Post-Pandemic Food Sourcing and Dietary Patterns among Urban Corner Store Customers in Baltimore, Maryland.

OBJECTIVE: Diet-related disease is rising, disproportionately affecting minority communities in which small food retail stores swamp supermarkets. Barriers to healthy food access were exacerbated by the pandemic. We examined the following: (1) individual- and household-level factors in a sample of Baltimore community members who regularly shop at corner stores and (2) how these factors are associated with indicators of dietary quality. DESIGN: Cross-sectional data were collected using an online survey to capture sociodemographics, anthropometrics, and food sourcing, spending, and consumption patterns. Concurrent quantitative and qualitative analyses were conducted in Stata 18 and ATLAS.ti. SETTING: This study was set in Baltimore, Maryland, USA. PARTICIPANTS: The participants included adults (n = 127) living or working in Baltimore who identified as regular customers of their neighborhood corner store. RESULTS: The respondents were majority Black and low-income, with a high prevalence of food insecurity (62.2%) and overweight/obesity (66.9%). Most (82.76%) shopped in their neighborhood corner store weekly. One-third (33.4%) of beverage calories were attributed to sugar-sweetened beverages, and few met the recommended servings for fruits and vegetables or fiber (27.2% and 10.4%, respectively). Being Black and not owning a home were associated with lower beverage and fiber intake, and not owning a home was also associated with lower fruit and vegetable intake. Food insecurity was associated with higher beverage intake, while WIC enrollment was associated with higher fruit and vegetable and fiber intakes. Open-ended responses contextualized post-pandemic food sourcing and consumption in this setting. CONCLUSIONS: This paper helps characterize the consumers of a complex urban food system. The findings will inform future strategies for consumer-engaged improvement of local food environments.

Machine learning-driven source identification and ecological risk prediction of heavy metal pollution in cultivated soils.

To overcome challenges in assessing the impact of environmental factors on heavy metal accumulation in soil due to limited comprehensive data, our study in Yangxin County, Hubei Province, China, analyzed 577 soil samples in combination with extensive big data. We used machine learning techniques, the potential ecological risk index, and the bivariate local Moran's index (BLMI) to predict Cr, Pb, Cd, As, and Hg concentrations in cultivated soil to assess ecological risks and identify pollution sources. The random forest model was selected for its superior performance among various machine learning models, and results indicated that heavy metal accumulation was substantially influenced by environmental factors such as climate, elevation, industrial activities, soil properties, railways, and population. Our ecological risk assessment highlighted areas of concern, where Cd and Hg were identified as the primary threats. BLMI was used to analyze spatial clustering and autocorrelation patterns between ecological risk and environmental factors, pinpointing areas that require targeted interventions. Additionally, redundancy analysis revealed the dynamics of heavy metal transfer to crops. This detailed approach mapped the spatial distribution of heavy metals, highlighted the ecological risks, identified their sources, and provided essential data for effective land management and pollution mitigation.

Design of a Mobile App Interface That Engages Community Members in a Food System Pilot Study.

Supermarkets are scarce in many under-resourced urban communities, and small independently owned retail stores often carry few fresh or healthy items. The Baltimore Urban food Distribution (BUD) mobile application (app) was previously developed to address supply-side challenges in moving healthy foods from local suppliers to retailers. In-app opportunities for consumers to indicate demand for these foods are crucial, but remain absent. We sought to understand community members' perspectives on the overall role, function and features of a proposed consumer-engagement module (BUDConnect) to expand the BUD app. A series of initial high-fidelity wireframe mockups were developed based on formative research. In-depth interviews (n = 20) were conducted and thematically analyzed using ATLAS.ti Web. Participants revealed a desire for real-time crowd-sourced information to navigate their food environments safely and effectively, functionality to help build community and social networks among store owners and their customers, opportunities to share positive reviews and ratings of store quality and offerings, and interoperability with existing apps. Rewards and referral systems resulting in the discounted purchasing of promoted healthy items were suggested to increase adoption and sustained app use. Wireframe mockups were further refined for future development and integration into the BUD app, the program and policy implications of which are discussed.

Fucoidan-hybrid hydroxyapatite nanoparticles promote the osteogenic differentiation of human periodontal ligament stem cells under inflammatory condition.

Inflammation-related bone defects often lead to poor osteogenesis. Therefore, it is crucial to reduce the inflammation response and promote the osteogenic differentiation of stem/progenitor cells to revitalize bone physiology. Here, a kind of hybrid nano-hydroxyapatite was prepared using the confined phosphate ion release method with the participation of fucoidan, a marine-sourced polysaccharide with anti-inflammation property. The physicochemical analyses confirmed that the fucoidan hybrid nano-hydroxyapatite (FC/n-HA) showed fine needle-like architectures. With a higher amount of fucoidan, the crystal size and crystallinity of the FC/n-HA reduced while the liquid dispersibility was improved. Cell experiences showed that FC/n-HA had an optimal cytocompatibility at concentration of 50 µg/mL. Moreover, the lipopolysaccharide-induced cellular inflammatory model with PDLSCs was established and used to evaluate the anti-inflammatory and osteogenic properties. For the 1%FC/n-HA group, the expression levels of TNF-α and IL-1ß were significantly reduced at 24 h, while the expression of alkaline phosphatase of PDLSCs was significantly promoted at days 3 and 7, and calcium precipitates was enhanced at 21 days. In this study, the FC/n-HA particles showed effective anti-inflammatory properties and facilitated osteogenic differentiation of PDLSCs, indicating which has potential application in treating bone defects associated with inflammation, such as periodontitis.

Machine Learning Model Based on Radiomics for Preoperative Differentiation of Jaw Cystic Lesions.

OBJECTIVE: This study aims to use machine learning techniques together with radiomics methods to build a preoperative predictive diagnostic model from spiral computed tomography (CT) images. The model is intended for the differential diagnosis of common jaw cystic lesions. STUDY DESIGN: Retrospective, case-control study. SETTING: This retrospective study was conducted at Sun Yat-sen Memorial Hospital of Sun Yat-sen University (Guangzhou, Guangdong, China). All the data used to build the predictive diagnostic model were collected from 160 patients, who were treated at the Department of Oral and Maxillofacial Surgery at Sun Yat-sen Memorial Hospital of Sun Yat-sen University between 2019 and 2023. METHODS: We included a total of 160 patients in this study. We extracted 107 radiomic features from each patient's CT scan images. After a feature selection process, we chose 15 of these radiomic features to construct the predictive diagnostic model. RESULTS: Among the preoperative predictive diagnostic models built using 3 different machine learning methods (support vector machine, random forest [RF], and multivariate logistic regression), the RF model showed the best predictive performance. It demonstrated a sensitivity of 0.923, a specificity of 0.643, an accuracy of 0.825, and an area under the receiver operating characteristic curve of 0.810. CONCLUSION: The preoperative predictive model, based on spiral CT radiomics and machine learning algorithms, shows promising differential diagnostic capabilities. For common jaw cystic lesions, this predictive model has potential clinical application value, providing a scientific reference for treatment decisions.

Chronic trans fatty acid consumption shortens lifespan in male Drosophila melanogaster on a high-sugar and high-fat diet.

Aging entails the progressive decline in the body's self-regulation and functionality over time. Notably, obesity and aging exhibit parallel phenotypes, with obesity further accelerating the aging process across multiple dimensions and diminishing lifespan. In this study, we explored the impact of trans fatty acid (TFA) consumption on the overall health and lifespan of male Drosophila melanogaster under an isocaloric high-sugar and high-fat diet. Our results indicate that TFA intake results in a shortened lifespan, elevated body weight, and increased triglyceride levels in flies fed a high-sugar and high-fat diet with equivalent caloric intake. Additionally, TFA exposure induces oxidative stress, locomotor deficits, and damage to the intestinal barrier in flies. Collectively, chronic TFA consumption expedites the aging process and reduces the lifespan of male Drosophila melanogaster. These results contribute supplementary evidence regarding the adverse health effects associated with TFAs.

A Turing test of whether AI chatbots are behaviorally similar to humans.

We administer a Turing test to AI chatbots. We examine how chatbots behave in a suite of classic behavioral games that are designed to elicit characteristics such as trust, fairness, risk-aversion, cooperation, etc., as well as how they respond to a traditional Big-5 psychological survey that measures personality traits. ChatGPT-4 exhibits behavioral and personality traits that are statistically indistinguishable from a random human from tens of thousands of human subjects from more than 50 countries. Chatbots also modify their behavior based on previous experience and contexts "as if" they were learning from the interactions and change their behavior in response to different framings of the same strategic situation. Their behaviors are often distinct from average and modal human behaviors, in which case they tend to behave on the more altruistic and cooperative end of the distribution. We estimate that they act as if they are maximizing an average of their own and partner's payoffs.

Yo-Yo Dieting Delays Male Drosophila melanogaster Aging Through Enhanced Mitochondrial Function, Relative to Sustained High-Calorie Diet Feeding.

Weight regain subsequent to weight reduction resulting from dietary interventions represents a prevalent phenomenon recognized as "Yo-yo dieting." However, the impact of prolonged Yo-yo dieting on health, especially in relation to the aging process, remains poorly understood. This study aimed to investigate the influence of Yo-yo dieting on the aging process in male Drosophila melanogaster that have been exposed to a high-calorie (HC) diet. Fruit flies were fed with either a consistent HC diet or an alternating regimen of HC and low-calorie diets every 3 days (referred to as "Yo-yo dieting") for a total of 24 days. Biochemical assays were utilized to quantify levels of oxidative stress and activities of the mitochondrial respiratory chain complexes. The frozen section staining method was employed to assess the presence of lipid droplets, reactive oxygen species, cellular viability, and mitochondrial abundance in tissues. Additionally, we examined the expression of key regulators involved in mitochondrial dynamics and biogenic signaling pathways. Yo-yo dieting resulted in an extension of the fruit flies' lifespan, concomitant with reduced body weight, decreased body protein content, and lower triglyceride levels compared to continuous a HC diet feeding. Furthermore, Yo-yo dieting ameliorated impairments in motility and intestinal barrier function. Importantly, it improved mitochondrial function and upregulated the expression of essential mitochondrial fusion proteins, namely mitofusin 1 and mitofusin 2, optic atrophy 1, and peroxisome proliferator-activated receptor-γ coactivator-1α. Therefore, the practice of Yo-yo dieting extends the lifespan of fruit flies by modulating mitochondrial dynamics and the associated biogenic signaling pathways.

Modeling annotator preference and stochastic annotation error for medical image segmentation.

Manual annotation of medical images is highly subjective, leading to inevitable annotation biases. Deep learning models may surpass human performance on a variety of tasks, but they may also mimic or amplify these biases. Although we can have multiple annotators and fuse their annotations to reduce stochastic errors, we cannot use this strategy to handle the bias caused by annotators' preferences. In this paper, we highlight the issue of annotator-related biases on medical image segmentation tasks, and propose a Preference-involved Annotation Distribution Learning (PADL) framework to address it from the perspective of modeling an annotator's preference and stochastic errors so as to produce not only a meta segmentation but also the annotator-specific segmentation. Under this framework, a stochastic error modeling (SEM) module estimates the meta segmentation map and average stochastic error map, and a series of human preference modeling (HPM) modules estimate each annotator's segmentation and the corresponding stochastic error. We evaluated our PADL framework on two medical image benchmarks with different imaging modalities, which have been annotated by multiple medical professionals, and achieved promising performance on all five medical image segmentation tasks. Code is available at https://github.com/Merrical/PADL.

ReFs: A hybrid pre-training paradigm for 3D medical image segmentation.

Self-supervised learning (SSL) has achieved remarkable progress in medical image segmentation. The application of an SSL algorithm often follows a two-stage training process: using unlabeled data to perform label-free representation learning and fine-tuning the pre-trained model on the downstream tasks. One issue of this paradigm is that the SSL step is unaware of the downstream task, which may lead to sub-optimal feature representation for a target task. In this paper, we propose a hybrid pre-training paradigm that is driven by both self-supervised and supervised objectives. To achieve this, a supervised reference task is involved in self-supervised learning, aiming to improve the representation quality. Specifically, we employ the off-the-shelf medical image segmentation task as reference, and encourage learning a representation that (1) incurs low prediction loss on both SSL and reference tasks and (2) leads to a similar gradient when updating the feature extractor from either task. In this way, the reference task pilots SSL in the direction beneficial for the downstream segmentation. To this end, we propose a simple but effective gradient matching method to optimize the model towards a consistent direction, thus improving the compatibility of both SSL and supervised reference tasks. We call this hybrid pre-training paradigm reference-guided self-supervised learning (ReFs), and perform it on a large-scale unlabeled dataset and an additional reference dataset. The experimental results demonstrate its effectiveness on seven downstream medical image segmentation benchmarks.

The mechanism of high mobility group box-1 protein and its bidirectional regulation in tumors.

High mobility group protein box-1 (HMGB1) is a nonhistone chromatin-related protein widely found in eukaryotic cells. It is involved in the transcription, replication and repair of DNA to maintain nuclear homeostasis. It participates in cell growth, differentiation and signal transduction. Recent studies showed that HMGB1 has a bidirectional regulatory effect on tumors by regulating TLR4/MYD88/NF-κB and RAGE/AMPK/mTOR signaling pathways. On one hand, it is highly expressed in a variety of tumors, promoting tumor proliferation and invasion, whilst on the other hand, it induces autophagy and apoptosis of tumor cells and stimulates tumor-infiltrating lymphocytes to produce anti-tumor immune response. At present, HMGB1 could be used as a target to regulate the drug-resistance and prognostication in cancer. Clinical applications of HMGB1 in cancer need further in-depth studies.

Prediction of postoperative recurrence of oral cancer by artificial intelligence model: Multilayer perceptron.

BACKGROUND: Postoperative recurrence of oral cancer is an important factor affecting the prognosis of patients. Artificial intelligence is used to establish a machine learning model to predict the risk of postoperative recurrence of oral cancer. METHODS: The information of 387 patients with postoperative oral cancer were collected to establish the multilayer perceptron (MLP) model. The comprehensive variable model was compared with the characteristic variable model, and the MLP model was compared with other models to evaluate the sensitivity of different models in the prediction of postoperative recurrence of oral cancer. RESULTS: The overall performance of the MLP model under comprehensive variable input was the best. CONCLUSION: The MLP model has good sensitivity to predict postoperative recurrence of oral cancer, and the predictive model with variable input training is better than that with characteristic variable input.

Learning From Partially Labeled Data for Multi-Organ and Tumor Segmentation.

Medical image benchmarks for the segmentation of organs and tumors suffer from the partially labeling issue due to its intensive cost of labor and expertise. Current mainstream approaches follow the practice of one network solving one task. With this pipeline, not only the performance is limited by the typically small dataset of a single task, but also the computation cost linearly increases with the number of tasks. To address this, we propose a Transformer based dynamic on-demand network (TransDoDNet) that learns to segment organs and tumors on multiple partially labeled datasets. Specifically, TransDoDNet has a hybrid backbone that is composed of the convolutional neural network and Transformer. A dynamic head enables the network to accomplish multiple segmentation tasks flexibly. Unlike existing approaches that fix kernels after training, the kernels in the dynamic head are generated adaptively by the Transformer, which employs the self-attention mechanism to model long-range organ-wise dependencies and decodes the organ embedding that can represent each organ. We create a large-scale partially labeled Multi-Organ and Tumor Segmentation benchmark, termed MOTS, and demonstrate the superior performance of our TransDoDNet over other competitors on seven organ and tumor segmentation tasks. This study also provides a general 3D medical image segmentation model, which has been pre-trained on the large-scale MOTS benchmark and has demonstrated advanced performance over current predominant self-supervised learning methods.

Computed tomography (CT) characteristics and pathologic basis of ciliated muconodular papillary tumors of the lung.

Background: Ciliated muconodular papillary tumor (CMPT) is a rare pulmonary tumor with papillary architecture. Most studies have focused on the clinicopathological features of CMPT, while computed tomography (CT) characteristics have rarely been systematically described. Methods: A cohort of 27 patients with surgically resected CMPT were identified. Clinical and demographic features were recorded. Preoperative CT images of the CMPTs and the corresponding histopathological basis were also retrospectively analyzed. Results: All of the tumors appeared as solitary nodules. Pure ground glass, part-solid nodules and solid nodules were detected in 2/27 (7.4%), 17/27 (63.0%), and 8/27 (29.6%) patients, respectively. Twenty-one tumors (77.8%) were located in the lower lobe. The average tumor size was 1.21±0.74 (range, 0.44-3.46) cm. Eighteen (66.7%) of the 27 patients had tumors with well-defined margins and lobulated contours. Fifteen patients (55.6%) had air bronchograms in the tumor, and 19 patients (70.4%) had air-containing space. There were two patients whose tumor size was enlarged and accompanied by an increase in solid components, and one patient simply had an increase in tumor size at the preoperative follow-up duration. Notably, one patient with solid tumor components was finally diagnosed with CMPT accompanied by adenocarcinoma. Conclusions: CMPTs of the lung mostly manifest as solitary, lobulated, well-defined tumors with air-containing spaces on CT and often occur in the periphery of the pulmonary lower lobe. When CT findings meet these criteria, the possibility of CMPT should be considered. Additionally, CMPT can coexist with adenocarcinoma. Further investigation will contribute significantly to the biological properties of CMPT and its relationship to the potential for malignant transformation.

Human gingival mesenchymal stem cell-derived exosomes cross-regulate the Wnt/ß-catenin and NF-κB signalling pathways in the periodontal inflammation microenvironment.

AIM: To examine the immunomodulatory effect of exosomes originating from gingival mesenchymal stem cells (GMSC-Exo) on periodontal bone regeneration and its role in the regulation of the nuclear-factor kappaB (NF-κB) and Wnt/ß-catenin pathways in the periodontal inflammatory microenvironment. MATERIALS AND METHODS: First, periodontal ligament stem cells (PDLSCs) were treated with GMSC-Exo or Porphyromonas gingivalis-derived lipopolysaccharide (P.g-LPS) in vitro. Quantitative real-time PCR (qRT-PCR) and western blot were carried out to detect the expressions of osteogenic differentiation-related factors in cells. Further, PDLSCs were treated with P.g-LPS or inhibitors. The expression of NF-κB pathway-related factors as well as of Wnt/ß-catenin pathway-related factors were detected by qRT-PCR and western blot. RESULTS: GMSC-Exo treatment promoted the expression of osteogenic differentiation-related factors within PDLSCs in both normal and inflammatory environments. Further investigations showed that GMSC-Exo could also inhibit the P.g-LPS-induced activation of the NF-κB pathway, leading to the up-regulation of the Wnt/ß-catenin pathway. When the Wnt/ß-catenin signalling was blocked, the inhibitory effect of GMSC-Exo on the NF-κB pathway was abolished. CONCLUSIONS: GMSC-Exo could promote the osteogenic differentiation of PDLSCs. There could be mutually exclusive regulatory roles between the NF-κB and Wnt/ß-catenin signalling pathways in a periodontal inflammatory environment. GMSC-Exo exhibited an effective cross-regulation ability for both pathways.