Attention-enhanced multiscale feature fusion network for pancreas and tumor segmentation.

BACKGROUND: Accurate pancreas and pancreatic tumor segmentation from abdominal scans is crucial for diagnosing and treating pancreatic diseases. Automated and reliable segmentation algorithms are highly desirable in both clinical practice and research. PURPOSE: Segmenting the pancreas and tumors is challenging due to their low contrast, irregular morphologies, and variable anatomical locations. Additionally, the substantial difference in size between the pancreas and small tumors makes this task difficult. This paper proposes an attention-enhanced multiscale feature fusion network (AMFF-Net) to address these issues via 3D attention and multiscale context fusion methods. METHODS: First, to prevent missed segmentation of tumors, we design the residual depthwise attention modules (RDAMs) to extract global features by expanding receptive fields of shallow layers in the encoder. Second, hybrid transformer modules (HTMs) are proposed to model deep semantic features and suppress irrelevant regions while highlighting critical anatomical characteristics. Additionally, the multiscale feature fusion module (MFFM) fuses adjacent top and bottom scale semantic features to address the size imbalance issue. RESULTS: The proposed AMFF-Net was evaluated on the public MSD dataset, achieving 82.12% DSC for pancreas and 57.00% for tumors. It also demonstrated effective segmentation performance on the NIH and private datasets, outperforming previous State-Of-The-Art (SOTA) methods. Ablation studies verify the effectiveness of RDAMs, HTMs, and MFFM. CONCLUSIONS: We propose an effective deep learning network for pancreas and tumor segmentation from abdominal CT scans. The proposed modules can better leverage global dependencies and semantic information and achieve significantly higher accuracy than the previous SOTA methods.

Two-dimensional crystalline platinum oxide.

Platinum (Pt) oxides are vital catalysts in numerous reactions, but research indicates that they decompose at high temperatures, limiting their use in high-temperature applications. In this study, we identify a two-dimensional (2D) crystalline Pt oxide with remarkable thermal stability (1,200 K under nitrogen dioxide) using a suite of in situ methods. This 2D Pt oxide, characterized by a honeycomb lattice of Pt atoms encased between dual oxygen layers forming a six-pointed star structure, exhibits minimized in-plane stress and enhanced vertical bonding due to its unique structure, as revealed by theoretical simulations. These features contribute to its high thermal stability. Multiscale in situ observations trace the formation of this 2D Pt oxide from α-PtO2, providing insights into its formation mechanism from the atomic to the millimetre scale. This 2D Pt oxide with outstanding thermal stability and distinct surface electronic structure subverts the previously held notion that Pt oxides do not exist at high temperatures and can also present unique catalytic capabilities. This work expands our understanding of Pt oxidation species and sheds light on the oxidative and catalytic behaviours of Pt oxide in high-temperature settings.

Exploring the Facet-Dependent Structural Evolution of Pt/CeO2 Catalysts Induced by Typical Pretreatments for CO Oxidation.

Atomic-scale insights into the interactions between metals and supports play a crucial role in optimizing catalyst design, understanding catalytic mechanisms, and enhancing chemical conversion processes. The effects of oxide support on the dynamic behavior of supported metal species during pretreatments or reactions have been attracting a lot of attention; however, very less systematic integrations are carried out experimentally using real catalysts. In this study, we here utilized facet-controlled CeO2 as examples to explore their influence on the supported Pt species (1.0 wt %) during the reducing and oxidizing pretreatments that are typically applied in heterogeneous catalysts. By employing a combination of microscopy, spectroscopy, and first-principles calculations, it is demonstrated that the exposed crystal facets of CeO2 govern the evolution behavior of supported Pt species under different environmental conditions. This leads to distinct local coordinations and charge states of the Pt species, which directly influence the catalytic reactivity and can be leveraged to control the catalytic performance for CO oxidation reactions.

Bubble-water/catalyst triphase interface microenvironment accelerates photocatalytic OER via optimizing semi-hydrophobic OH radical.

Photocatalytic water splitting (PWS) as the holy grail reaction for solar-to-chemical energy conversion is challenged by sluggish oxygen evolution reaction (OER) at water/catalyst interface. Experimental evidence interestingly shows that temperature can significantly accelerate OER, but the atomic-level mechanism remains elusive in both experiment and theory. In contrast to the traditional Arrhenius-type temperature dependence, we quantitatively prove for the first time that the temperature-induced interface microenvironment variation, particularly the formation of bubble-water/TiO2(110) triphase interface, has a drastic influence on optimizing the OER kinetics. We demonstrate that liquid-vapor coexistence state creates a disordered and loose hydrogen-bond network while preserving the proton transfer channel, which greatly facilitates the formation of semi-hydrophobic •OH radical and O-O coupling, thereby accelerating OER. Furthermore, we propose that adding a hydrophobic substance onto TiO2(110) can manipulate the local microenvironment to enhance OER without additional thermal energy input. This result could open new possibilities for PWS catalyst design.

A Multi-Resolution Denoising Method for Low-Dose CT Based on the Reconstruction of Wavelet High-Frequency Channel.

Computed tomography (CT) is widely applied in contemporary clinic. Due to the radiation risks carried by X-rays, the imaging and post-processing methods of low-dose CT (LDCT) become popular topics in academia and industrial community. Generally, LDCT presents strong noise and artifacts, while existing algorithms cannot completely overcome the blurring effects and meantime reduce the noise. The proposed method enables CT extend to independent frequency channels by wavelet transformation, then two separate networks are established for low-frequency denoising and high-frequency reconstruction. The clean signals from high-frequency channel are reconstructed through channel translation, which is essentially effective in preserving detailed structures. The public dataset from Mayo Clinic was used for model training and testing. The experiments showed that the proposed method achieves a better quantitative result (PSNR: 37.42dB, SSIM: 0.8990) and details recovery visually, which demonstrates our framework can better restore the detailed features while significantly suppressing the noise.

Multimodality Fusion Method Based on Multiview Subspace Clustering for Pulmonary Embolism Diagnosis.

Pulmonary embolism (PE) is an important clinical disorder that will result in lung tissue damage or low blood oxygen levels, which need early diagnosis and timely treatment. While computed tomographic pulmonary angiography (CTPA) is the gold standard to diagnose PE, previous studies have verified the effectiveness of combing CTPA and EMR data in computer-aided PE detection or diagnosis. In this paper, we proposed a multimodality fusion method based on multi-view subspace clustering guided feature selection (MSCUFS). The extracted high-dimensional image and EMR features are firstly selected and fused by the MSCUFS, and then are feed into different machine learning models with different fusion strategy to construct the PE classifier. The experiment results showed that the joint fusion strategy with MSCUFS achieved best AUROC of 0.947, surpassing other early fusion and late fusion models. The comparison between single modality and multimodality also illustrated the effectiveness of the proposed method.

Dual-Attention Model Fusing CNN and Transformer for Pancreas Segmentation.

Pancreatic cancer is a highly malignant cancer of the digestive tract and is rapidly progressing and spreading clinically. Automatic and accurate pancreatic tissue segmentation in abdominal CT images is essential for the early diagnosis of pancreatic-related diseases. It is challenging that the pancreas is small in size and complex in morphology. To address this problem, we propose a dual-attention model fusing CNN and Transformer to effectively activate pancreas-related features expression. The CNN structure weights the importance of pancreas-related features at the channel level and weakens the background information. Transformer feature aggregation module constructs spatial correlations among long-distance pixels from a global perspective. This study is validated on the NIH-TCIA dataset and achieved a mean Dice Similarity Coefficient of 85.82%, which is outperforming than the state-of-the-art methods. The visualization of surface distance also demonstrates the effective segmentation of pancreas boundary details by the proposed model.

Position Prior Attention Network for Pancreas Tumor Segmentation.

Segmentation of pancreatic tumors on CT images is essential for the diagnosis and treatment of pancreatic cancer. However, low contrast between the pancreas and the tumor, as well as variable tumor shape and position, makes segmentation challenging. To solve the problem, we propose a Position Prior Attention Network (PPANet) with a pseudo segmentation generation module (PSGM) and a position prior attention module (PPAM). PSGM and PPAM maps pancreatic and tumor pseudo segmentation to latent space to generate position prior attention map and supervises location classification. The proposed method is evaluated on pancreatic patient data collected from local hospital and the experimental results demonstrate that our method can significantly improve the tumor segmentation results by introducing the position information in the training phase.

Whole-Liver Based Deep Learning for Preoperatively Predicting Overall Survival in Patients with Hepatocellular Carcinoma.

Survival prediction is crucial for treatment decision making in hepatocellular carcinoma (HCC). We aimed to build a fully automated artificial intelligence system (FAIS) that mines whole-liver information to predict overall survival of HCC. We included 215 patients with preoperative contrast-enhance CT imaging and received curative resection from a hospital in China. The cohort was randomly split into developing and testing subcohorts. The FAIS was constructed with convolutional layers and full-connected layers. Cox regression loss was used for training. Models based on clinical and/or tumor-based radiomics features were built for comparison. The FAIS achieved C-indices of 0.81 and 0.72 for the developing and testing sets, outperforming all the other three models. In conclusion, our study suggest that more important information could be mined from whole liver instead of only the tumor. Our whole-liver based FAIS provides a non-invasive and efficient overall survival prediction tool for HCC before the surgery.

Mining Whole-liver Information with Deep Learning for Preoperatively Predicting HCC Recurrence-free Survival.

Hepatocellular carcinoma (HCC) is globally a leading cause of cancer death. Non-invasive pre-operative prediction of HCC recurrence-free survival (RFS) after resection is essential but remains challenging. Previous models based on medical imaging focus only on tumor area while neglecting the whole liver condition. In fact, HCC patients usually suffer from chronic liver diseases which also hamper the patient survival. This work aims to develop a novel convolutional neural network (CNN) to mine whole-liver information from contrast-enhanced computed tomography (CECT) to predict RFS after hepatic resection in HCC. Our proposed RFSNet takes liver regions from CECT as input, and outputs a risk score for each patient. Cox proportional-hazards loss was applied for model training. A total of 215 patients with primary HCC and treated with hepatic resection were included for analysis. Patients were randomly split into developing subcohort and testing subcohort by 4:1. The developing subcohort was further split into the training subcohort and validation subcohort for model training. Baseline models were built with tumor region, radiomics features and/or clinical features the same as previous tumor-based approaches. Results showed that RFSNet achieved the best performance with concordance-indinces (CIs) of 0.88 and 0.65 for the developing and testing subcohorts, respectively. Adding clinical features did not improve RFSNet. Our findings suggest that the proposed RFSNet based on whole liver is able to extract more valuable information concerning RFS prognosis compared to features from only tumor and the clinical indicators.

Siamese Delta Network with Multimodality Fusion for Systemic Treatment Response Prediction in Pancreatic Cancer.

Systemic treatment is a main way for pancreas cancer patients that are ineligible for surgery. A subgroup of patients showed good response to systemic treatment and the rest received limited benefits. CT images provide a non-invasive way to assess the treatment response. Alternative non-image methods include radiology analysis, tumor marker analysis and combination analysis. To combine the image and non-image data, we propose the Siamese Delta Network with Multimodality Fusion (SDN-MF) to predict systemic treatment response in an end-to-end way. First, a Siamese Delta Network (SDN) is designed to process pre-treatment and pre-surgery CT images and get the image feature changes to predict response. Then, patients' characteristics from EMR and alternative analysis results forms non-image data, which is incorporated into SDN with a multimodality fusion (MF) module. The proposed SDN-MF is evaluated on a private dataset and achieves average AUC value of 0.883 with five cross-validation. Comparison among image-only, non-image-only, and fusion models verifies the superior of multimodality model in predicting systemic treatment response of pancreas cancer patients.

Partial Label Multi-organ Segmentation based on Local Feature Enhancement.

The automatic segmentation of abdominal organs from CT images is essential for surgical planning of abdominal diseases. However, each medical institution only annotates some organs according to its own clinical practice. This brings the partial annotation problem to multi-center abdominal multi-organ segmentation. To address this issue, we introduce a 3D local feature enhanced multi-head segmentation network for multi-organ segmentation of abdominal regions in multiple partially labeled datasets. More specifically, our proposed architecture consists of two branches, the global branch with 3D Transformer and U-Net fusion named 3D TransUNet as the backbone, and the local 3D U-Net branch that provides additional abdominal organ structure information to the global branch to generate more accurate segmentation results. We evaluate our method on four publicly available CT datasets with four different partial label. Our experiments show that the proposed approach provides better accuracy and robustness, with 93.01% average Dice-score-coefficient (DSC) and 3.489 mm Hausdorff Distance (HD) outperforming three existing state-of-the-art methods.

Prediction of New-Onset Diabetes after Pancreatectomy with Subspace Clustering based Multi-view Feature Selection.

The pancreas plays an important role in glucose metabolism, and developing diabetes or long-term glucose metabolism disturbance may be a prevalent sequela after pancreatectomy. Nevertheless, relative factors of new-onset diabetes after pancreatectomy stay unclear. Radiomics analysis is potential to identify image markers for disease prediction or prognosis. Meanwhile, combination of imaging and electronic medical record (EMR) showed superior performance than imaging or EMR alone in previous studies. One critical step is to identity predictors from high-dimensional features, and it is even more challenging to select and fuse imaging and EMR features. In this work, we develop a radiomics pipeline to assess postoperative new-onset diabetes risk of patients undergoing distal pancreatectomy. Specifically, we extract multiscale image features with 3D wavelet transformation, and include patients' characteristics, body composition and pancreas volume information as clinical features. Then, we propose a multi-view subspace clustering guided feature selection method (MSCUFS) for the selection and fusion of image and clinical features. Finally, a prediction model is constructed with classical machine learning classifier. Experimental results on an established distal pancreatectomy cohort showed that the SVM model with combined imaging and EMR features demonstrated good discrimination, with an AUC value of 0.824, which improved the model with image features alone by 0.037 AUC. Compared with state-of-the-art feature selection methods, the proposed MSCUFS has superior performance in fusing image and clinical features.

Mass Transport and Electron Transfer at the Electrochemical-Confined Interface.

Single entity measurements based on the stochastic collision electrochemistry provide a promising and versatile means to study single molecules, single particles, single droplets, etc. Conceptually, mass transport and electron transfer are the two main processes at the electrochemically confined interface that underpin the most transient electrochemical responses resulting from the stochastic and discrete behaviors of single entities at the microscopic scale. This perspective demonstrates how to achieve controllable stochastic collision electrochemistry by effectively altering the two processes. Future challenges and opportunities for stochastic collision electrochemistry are also highlighted.

Multiscale unsupervised domain adaptation for automatic pancreas segmentation in CT volumes using adversarial learning.

PURPOSE: Computer-aided automatic pancreas segmentation is essential for early diagnosis and treatment of pancreatic diseases. However, the annotation of pancreas images requires professional doctors and considerable expenditure. Due to imaging differences among various institution population, scanning devices, imaging protocols, and so on, significant degradation in the performance of model inference results is prone to occur when models trained with domain-specific (usually institution-specific) datasets are directly applied to new (other centers/institutions) domain data. In this paper, we propose a novel unsupervised domain adaptation method based on adversarial learning to address pancreas segmentation challenges with the lack of annotations and domain shift interference. METHODS: A 3D semantic segmentation model with attention module and residual module is designed as the backbone pancreas segmentation model. In both segmentation model and domain adaptation discriminator network, a multiscale progressively weighted structure is introduced to acquire different field of views. Features of labeled data and unlabeled data are fed in pairs into the proposed multiscale discriminator to learn domain-specific characteristics. Then the unlabeled data features with pseudodomain label are fed to the discriminator to acquire domain-ambiguous information. With this adversarial learning strategy, the performance of the segmentation network is enhanced to segment unseen unlabeled data. RESULTS: Experiments were conducted on two public annotated datasets as source datasets, respectively, and one private dataset as target dataset, where annotations were not used for the training process but only for evaluation. The 3D segmentation model achieves comparative performance with state-of-the-art pancreas segmentation methods on source domain. After implementing our domain adaptation architecture, the average dice similarity coefficient (DSC) of the segmentation model trained on the NIH-TCIA source dataset increases from 58.79% to 72.73% on the local hospital dataset, while the performance of the target domain segmentation model transferred from the medical segmentation decathlon (MSD) source dataset rises from 62.34% to 71.17%. CONCLUSIONS: Correlations of features across data domains are utilized to train the pancreas segmentation model on unlabeled data domain, improving the generalization of the model. Our results demonstrate that the proposed method enables the segmentation model to make meaningful segmentation for unseen data of the training set. In the future, the proposed method has the potential to apply segmentation model trained on public dataset to clinical unannotated CT images from local hospital, effectively assisting radiologists in clinical practice.

Hydrogen Spillover-Bridged Volmer/Tafel Processes Enabling Ampere-Level Current Density Alkaline Hydrogen Evolution Reaction under Low Overpotential.

Water-alkaline electrolysis holds a great promise for industry-scale hydrogen production but is hindered by the lack of enabling hydrogen evolution reaction electrocatalysts to operate at ampere-level current densities under low overpotentials. Here, we report the use of hydrogen spillover-bridged water dissociation/hydrogen formation processes occurring at the synergistically hybridized Ni3S2/Cr2S3 sites to incapacitate the inhibition effect of high-current-density-induced high hydrogen coverage at the water dissociation site and concurrently promote Volmer/Tafel processes. The mechanistic insights critically important to enable ampere-level current density operation are depicted from the experimental and theoretical studies. The Volmer process is drastically boosted by the strong H2O adsorption at Cr5c sites of Cr2S3, the efficient H2O* dissociation via a heterolytic cleavage process (Cr5c-H2O* + S3c(#) → Cr5c-OH* + S3c-H#) on the Cr5c/S3c sites in Cr2S3, and the rapid desorption of OH* from Cr5c sites of Cr2S3 via a new water-assisted desorption mechanism (Cr5c-OH* + H2O(aq) → Cr5c-H2O* + OH-(aq)), while the efficient Tafel process is achieved through hydrogen spillover to rapidly transfer H# from the synergistically located H-rich site (Cr2S3) to the H-deficient site (Ni3S2) with excellent hydrogen formation activity. As a result, the hybridized Ni3S2/Cr2S3 electrocatalyst can readily achieve a current density of 3.5 A cm-2 under an overpotential of 251 ± 3 mV in 1.0 M KOH electrolyte. The concept exemplified in this work provides a useful means to address the shortfalls of ampere-level current-density-tolerant Hydrogen evolution reaction (HER) electrocatalysts.

Understanding the Dynamic Potential Distribution at the Electrode Interface by Stochastic Collision Electrochemistry.

The potential distribution at the electrode interface is a core factor in electrochemistry, and it is usually treated by the classic Gouy-Chapman-Stern (G-C-S) model. Yet the G-C-S model is not applicable to nanosized particles collision electrochemistry as it describes steady-state electrode potential distribution. Additionally, the effect of single nanoparticles (NPs) on potential should not be neglected because the size of a NP is comparable to that of an electrode. Herein, a theoretical model termed as Metal-Solution-Metal Nanoparticle (M-S-MNP) is proposed to reveal the dynamic electrode potential distribution at the single-nanoparticle level. An explicit equation is provided to describe the size/distance-dependent potential distribution in single NPs stochastic collision electrochemistry, showing the potential distribution is influenced by the NPs. Agreement between experiments and simulations indicates the potential roles of the M-S-MNP model in understanding the charge transfer process at the nanoscale.

Potential Role of Methanogens in Microbial Reductive Dechlorination of Organic Chlorinated Pollutants In Situ.

Previous studies often attribute microbial reductive dechlorination to organohalide-respiring bacteria (OHRB) or cometabolic dechlorination bacteria (CORB). Even though methanogenesis frequently occurs during dechlorination of organic chlorinated pollutants (OCPs) in situ, the underestimated effect of methanogens and their interactions with dechlorinators remains unknown. We investigated the association between dechlorination and methanogenesis, as well as the performance of methanogens involved in reductive dechlorination, through the use of meta-analysis, incubation experiment, untargeted metabolomic analysis, and thermodynamic modeling approaches. The meta-analysis indicated that methanogenesis is largely synchronously associated with OCP dechlorination, that OHRB are not the sole degradation engineers that maintain OCP bioremediation, and that methanogens are fundamentally needed to sustain microenvironment functional balance. Laboratory results further confirmed that Methanosarcina barkeri (M. barkeri) promotes the dechlorination of γ-hexachlorocyclohexane (γ-HCH). Untargeted metabolomic analysis revealed that the application of γ-HCH upregulated the metabolic functioning of chlorocyclohexane and chlorobenzene degradation in M. barkeri, further confirming that M. barkeri potentially possesses an auxiliary dechlorination function. Finally, quantum analysis based on density functional theory (DFT) indicated that the methanogenic coenzyme F430 significantly reduces the activation barrier to dechlorination. Collectively, this work suggests that methanogens are highly involved in microbial reductive dechlorination at OCP-contaminated sites and may even directly favor OCP degradation.

SSIA: A sensitivity-supervised interlock algorithm for high-performance microkinetic solving.

Microkinetic modeling has drawn increasing attention for quantitatively analyzing catalytic networks in recent decades, in which the speed and stability of the solver play a crucial role. However, for the multi-step complex systems with a wide variation of rate constants, the often encountered stiff problem leads to the low success rate and high computational cost in the numerical solution. Here, we report a new efficient sensitivity-supervised interlock algorithm (SSIA), which enables us to solve the steady state of heterogeneous catalytic systems in the microkinetic modeling with a 100% success rate. In SSIA, we introduce the coverage sensitivity of surface intermediates to monitor the low-precision time-integration of ordinary differential equations, through which a quasi-steady-state is located. Further optimized by the high-precision damped Newton's method, this quasi-steady-state can converge with a low computational cost. Besides, to simulate the large differences (usually by orders of magnitude) among the practical coverages of different intermediates, we propose the initial coverages in SSIA to be generated in exponential space, which allows a larger and more realistic search scope. On examining three representative catalytic models, we demonstrate that SSIA is superior in both speed and robustness compared with its traditional counterparts. This efficient algorithm can be promisingly applied in existing microkinetic solvers to achieve large-scale modeling of stiff catalytic networks.

CATKINAS: A large-scale catalytic microkinetic analysis software for mechanism auto-analysis and catalyst screening.

As an effective method to analyze complex catalytic reaction networks, microkinetic modeling is gaining increasing popularity in the catalytic activity evaluation and rational design of heterogeneous catalysts. An automated simulator with stable and reliable performance is especially useful and in great request. Here we introduce the CATKINAS package developed for large-scale microkinetic modeling and analysis. Featuring with a multilevel solver and a multifunctional analyzer, CATKINAS can provide both accurate solutions and various quantitative and automatic analysis for a wide range of catalytic systems. The structure and the basic workflow are overviewed with the multilevel solver particularly illustrated. Also, we take the CO methanation reaction as an example to illustrate the application and efficiency of the CATKINAS package.