Using machine-learning models to predict extubation failure in neonates with bronchopulmonary dysplasia.

AIM: To develop a decision-support tool for predicting extubation failure (EF) in neonates with bronchopulmonary dysplasia (BPD) using a set of machine-learning algorithms. METHODS: A dataset of 284 BPD neonates on mechanical ventilation was used to develop predictive models via machine-learning algorithms, including extreme gradient boosting (XGBoost), random forest, support vector machine, naïve Bayes, logistic regression, and k-nearest neighbor. The top three models were assessed by the area under the receiver operating characteristic curve (AUC), and their performance was tested by decision curve analysis (DCA). Confusion matrix was used to show the high performance of the best model. The importance matrix plot and SHapley Additive exPlanations values were calculated to evaluate the feature importance and visualize the results. The nomogram and clinical impact curves were used to validate the final model. RESULTS: According to the AUC values and DCA results, the XGboost model performed best (AUC = 0.873, sensitivity = 0.896, specificity = 0.838). The nomogram and clinical impact curve verified that the XGBoost model possessed a significant predictive value. The following were predictive factors for EF: pO2, hemoglobin, mechanical ventilation (MV) rate, pH, Apgar score at 5 min, FiO2, C-reactive protein, Apgar score at 1 min, red blood cell count, PIP, gestational age, highest FiO2 at the first 24 h, heart rate, birth weight, pCO2. Further, pO2, hemoglobin, and MV rate were the three most important factors for predicting EF. CONCLUSIONS: The present study indicated that the XGBoost model was significant in predicting EF in BPD neonates with mechanical ventilation, which is helpful in determining the right extubation time among neonates with BPD to reduce the occurrence of complications.

Life makes cities greener: The impact of dual-policy of China in urban transformation on residents' green lifestyles.

Promoting the formation of the green lifestyle (GL) is a crucial step in achieving comprehensive green transformation of urban economic and social development. The widespread adoption of GL is influenced by various environmental regulations. Previous research mainly focused on the impact of individual policies on GL from the single policy perspective. The mechanisms of the combined effects of policies have not been thoroughly explored, particularly the contributions of each policy during periods of overlap. This paper takes the dual-policy of the New-type Urbanization Policy (NUP) and Smart City Policy (SCP) in China as an example. It employs panel data collected from 271 cities in China during 2007-2019 and establishes a multi-period difference-in-difference model to identify the combined effects of the dual-policy on residents' GL. Additionally, the Shapley value decomposition method is utilized to identify the contribution magnitude of each policy when they act simultaneously. The following conclusions are yielded. Firstly, the combined effects of dual-policy are more effective than a single policy in influencing residents' GL. Secondly, the Shapley value decomposition method reveals that when both policies are simultaneously implemented, SCP contributes a greater weight compared to NUP. Thirdly, the dual-policy can promote residents' adoption of GL through mechanisms such as green technological innovation, public participation in environmental protection, and the agglomeration of tertiary industries. Fourthly, the impact of dual-policy on residents' GL varies across different types and sizes of cities. This study attempts to unseal the "black box" of how the dual-policy influences residents' GL during the green transformation of cities in China, providing theoretical references for relevant urban policies in other countries and contributing to Chinese solutions and experience to global urban green development.

Effects of Different Farming Modes on Salmo trutta fario Growth and Intestinal Microbial Community.

The gut microbiota plays a pivotal role in upholding intestinal health, fostering intestinal development, fortifying organisms against pathogen intrusion, regulating nutrient absorption, and managing the body's lipid metabolism. However, the influence of different cultivation modes on the growth indices and intestinal microbes of Salmo trutta fario remains underexplored. In this study, we employed high-throughput sequencing and bioinformatics techniques to scrutinize the intestinal microbiota in three farming modes: traditional pond aquaculture (TPA), recirculating aquaculture (RA), and flow-through aquaculture (FTA). We aimed to assess the impact of different farming methods on the water environment and Salmo trutta fario's growth performance. Our findings revealed that the final weight and weight gain rate in the FTA model surpassed those in the other two. Substantial disparities were observed in the composition, relative abundance, and diversity of Salmo trutta fario gut microbiota under different aquaculture modes. Notably, the dominant genera of Salmo trutta fario gut microbiota varied across farming modes: for instance, in the FTA model, the most prevalent genera were SC-I-84 (7.34%), Subgroup_6 (9.93%), and UTCFX1 (6.71%), while, under RA farming, they were Bacteroidetes_vadinHA17 (10.61%), MBNT15 (7.09%), and Anaeromyxoactor (6.62%). In the TPA model, dominant genera in the gut microbiota included Anaeromyxobacter (8.72%), Bacteroidetes_vadinHA17 (8.30%), and Geobacter (12.54%). From a comparative standpoint, the genus-level composition of the gut microbiota in the RA and TPA models exhibited relative similarity. The gut microbiota in the FTA model showcased the most intricate functional diversity, while TPA farming displayed a more intricate interaction pattern with the gut microbiota. Transparency, pH, dissolved oxygen, conductivity, total dissolved solids, and temperature emerged as pivotal factors influencing Salmo trutta fario gut microbiota under diverse farming conditions. These research findings offer valuable scientific insights for fostering healthy aquaculture practices and disease prevention and control measures for Salmo trutta fario, holding substantial significance for the sustainable development of the cold-water fish industry in the Qinghai-Tibet Plateau.

Expression of lncRNAs in children with pancreaticobiliary maljunction: functional analysis and potential biomarkers.

Introduction: Pancreaticobiliary maljunction (PBM) leads to higher rates of complications, including cholangitis, pancreatitis, and malignancies. The aim of the present study was to investigate the expression profile of long non-coding RNAs (lncRNAs) and their potential role as biomarkers in children with pancreaticobiliary maljunction. Material and methods: The differential expression of lncRNAs and messenger RNA (mRNAs) from pediatric patients with pancreaticobiliary maljunction and control subjects was analyzed using a commercial microarray and later validated with qRT-PCR. The potential biological functions of differentially expressed genes were explored based on Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment. The ability of potential lncRNA biomarkers to predict pancreaticobiliary maljunction was assessed based on the area under the receiver operating characteristic curve (AUC). Results: There were 2915 mRNAs and 173 lncRNAs upregulated, and 2121 mRNAs and 316 lncRNAs downregulated in PBM cases compared to controls. The enriched Gene Ontology categories associated with differentially expressed mRNAs were extracellular matrix, extracellular region, and kinetochore. The most enriched Kyoto Encyclopedia pathway was protein digestion and absorption, which was associated with cancer and PI3K-Akt signaling. Analysis of cis- and trans-target genes predicted that a single lncRNA was able to regulate several mRNAs. The qRT-PCR results for NR_110876, NR_132344, XR_946886, and XR_002956345 were consistent with the microarray results, and the difference was statistically significant for NR_132344, XR_946886, and XR_002956345 (p < 0.05). AUC was significant only for XR_946886 (0.837, p < 0.001). Conclusions: Our results implicate lncRNAs in common bile duct pathogenesis in PBM, and they identify XR_946886 as a potential biomarker for the disease.

Iterative feedback-based models for image and video polyp segmentation.

Accurate segmentation of polyps in colonoscopy images has gained significant attention in recent years, given its crucial role in automated colorectal cancer diagnosis. Many existing deep learning-based methods follow a one-stage processing pipeline, often involving feature fusion across different levels or utilizing boundary-related attention mechanisms. Drawing on the success of applying Iterative Feedback Units (IFU) in image polyp segmentation, this paper proposes FlowICBNet by extending the IFU to the domain of video polyp segmentation. By harnessing the unique capabilities of IFU to propagate and refine past segmentation results, our method proves effective in mitigating challenges linked to the inherent limitations of endoscopic imaging, notably the presence of frequent camera shake and frame defocusing. Furthermore, in FlowICBNet, we introduce two pivotal modules: Reference Frame Selection (RFS) and Flow Guided Warping (FGW). These modules play a crucial role in filtering and selecting the most suitable historical reference frames for the task at hand. The experimental results on a large video polyp segmentation dataset demonstrate that our method can significantly outperform state-of-the-art methods by notable margins achieving an average metrics improvement of 7.5% on SUN-SEG-Easy and 7.4% on SUN-SEG-Hard. Our code is available at https://github.com/eraserNut/ICBNet.

Clinical evaluation of a novel digital microfluidic based point-of-care test platform for detection of SARS-Cov-2 and influenza A/B.

Respiratory pathogens, such as SARS-CoV-2 and influenza A/B, can cause severe illnesses in susceptible individuals. This research evaluated a novel digital microfluidic point-of-care testing platform designed to detect 23 pathogens, comparing its performance to conventional laboratory-based nucleic acid tests. The platform integrates nucleic acid extraction and amplification processes for rapid detection with only 2 min of hands-on time. Performance assays demonstrated that the platform has high sensitivity (87 %-100 %) and specificity (99 %-100 %) for the detection of the evaluated 3 viruses. Additionally, the platform can be adapted for the detection of other respiratory pathogens, aiding in the early diagnosis of respiratory diseases, identifying the source of an outbreak or epidemic, and curbing the spread of the disease.

Relation-Aware Heterogeneous Graph Network for Learning Intermodal Semantics in Textbook Question Answering.

Textbook question answering (TQA) task aims to infer answers for given questions from a multimodal context, including text and diagrams. The existing studies have aggregated intramodal semantics extracted from a single modality but have yet to capture the intermodal semantics between different modalities. A major challenge in learning intermodal semantics is maintaining lossless intramodal semantics while bridging the gap of semantics caused by heterogeneity. In this article, we propose an intermodal relation-aware heterogeneous graph network (IMR-HGN) to extract the intermodal semantics for TQA, which aggregates different modalities while learning features rather than representing them independently. First, we design a multidomain consistent representation (MDCR) to eliminate semantic gaps by capturing intermodal features while maintaining lossless intramodal semantics in multidomains. Furthermore, we present neighbor-based relation inpainting (NRI) to reduce semantic ambiguity via repairing fuzzy relations with correlations of relations. Finally, we propose hierarchical multisemantics aggregation (HMSA) to guarantee the completeness of semantics by aggregating features of nodes and relations with a reconstruction network (RN). Experimental results show that IMR-HGN could extract the intermodal semantics of answers, achieving a 2.16% improvement on the validation set of the TQA dataset and a 3.04% increase on the test set of the AI2D dataset.

Development of a machine learning model and nomogram to predict seizures in children with COVID-19: a two-center study.

OBJECTIVE: This study aimed to use machine learning to evaluate the risk factors of seizures and develop a model and nomogram to predict seizures in children with coronavirus disease 2019 (COVID-19). MATERIAL AND METHODS: A total of 519 children with COVID-19 were assessed to develop predictive models using machine learning algorithms, including extreme gradient boosting (XGBoost), random forest (RF) and logistic regression (LR). The performance of the models was assessed using area under the receiver operating characteristic curve (AUC) values. Importance matrix plot and SHapley Additive exPlanations (SHAP) values were calculated to evaluate feature importance and to show the visualization results. The nomogram and clinical impact curve were used to validate the final model. RESULTS: Two hundred and seventeen children with COVID-19 had seizures. According to the AUC, the RF model performed the best. Based on the SHAP values, the top three most important variables in the RF model were neutrophil percentage, cough and fever duration. The nomogram and clinical impact curve also verified that the RF model possessed significant predictive value. CONCLUSIONS: Our research indicates that the RF model demonstrates excellent performance in predicting seizures, and our novel nomogram can facilitate clinical decision-making and potentially offer benefit for clinicians to prevent and treat seizures in children with COVID-19.

Development of a CT-based radiomics-clinical model to diagnose acute pancreatitis on nonobvious findings on CT in children with pancreaticobiliary maljunction.

OBJECTIVES: Since neither abdominal pain nor pancreatic enzyme elevation is specific for acute pancreatitis (AP), the diagnosis of AP in patients with pancreaticobiliary maljunction (PBM) may be challenging when the pancreas appears normal or nonobvious on CT. This study aimed to develop a quantitative radiomics-based nomogram of pancreatic CT for identifying AP in children with PBM who have nonobvious findings on CT. METHODS: PBM patients with a diagnosis of AP evaluated at the Children's Hospital of Soochow University from June 2015 to October 2022 were retrospectively reviewed. The radiological features and clinical factors associated with AP were evaluated. Based on the selected variables, multivariate logistic regression was used to construct clinical, radiomics, and combined models. RESULTS: Two clinical parameters and 6 radiomics characteristics were chosen based on their significant association with AP, as demonstrated in the training (area under curve [AUC]: 0.767, 0.892) and validation (AUC: 0.757, 0.836) datasets. The radiomics-clinical nomogram demonstrated superior performance in both the training (AUC, 0.938) and validation (AUC, 0.864) datasets, exhibiting satisfactory calibration (P > .05). CONCLUSIONS: Our radiomics-based nomogram is an accurate, noninvasive diagnostic technique that can identify AP in children with PBM even when CT presentation is not obvious. ADVANCES IN KNOWLEDGE: This study extracted imaging features of nonobvious pancreatitis. Then it developed and evaluated a combined model with these features.

Hybrid Masked Image Modeling for 3D Medical Image Segmentation.

Masked image modeling (MIM) with transformer backbones has recently been exploited as a powerful self-supervised pre-training technique. The existing MIM methods adopt the strategy to mask random patches of the image and reconstruct the missing pixels, which only considers semantic information at a lower level, and causes a long pre-training time. This paper presents HybridMIM, a novel hybrid self-supervised learning method based on masked image modeling for 3D medical image segmentation. Specifically, we design a two-level masking hierarchy to specify which and how patches in sub-volumes are masked, effectively providing the constraints of higher level semantic information. Then we learn the semantic information of medical images at three levels, including: 1) partial region prediction to reconstruct key contents of the 3D image, which largely reduces the pre-training time burden (pixel-level); 2) patch-masking perception to learn the spatial relationship between the patches in each sub-volume (region-level); and 3) drop-out-based contrastive learning between samples within a mini-batch, which further improves the generalization ability of the framework (sample-level). The proposed framework is versatile to support both CNN and transformer as encoder backbones, and also enables to pre-train decoders for image segmentation. We conduct comprehensive experiments on five widely-used public medical image segmentation datasets, including BraTS2020, BTCV, MSD Liver, MSD Spleen, and BraTS2023. The experimental results show the clear superiority of HybridMIM against competing supervised methods, masked pre-training approaches, and other self-supervised methods, in terms of quantitative metrics, speed performance and qualitative observations.

Serum tRNA-derived fragments as potential biomarkers in children with acute intussusception.

Introduction: Pediatric intussusception is one of the most common causes of bowel obstruction in the pediatric population. Affected children have one section of the intestine sliding into the adjacent section. Intestinal ischemia-reperfusion injury (I/R) can occur during pediatric intussusception, and any delay in diagnosis or treatment can lead to loss of intestinal viability that requires bowel resection. The aim of the present study was to investigate whether transfer ribonucleic acid (tRNA)-derived fragments (tRFs) can serve as candidate biomarkers for pediatric intussusception. Material and methods: Using high-throughput sequencing technology, we identified differentially expressed tRFs, and ultimately selected three tRFs to establish a signature as a predictive biomarker of pediatric intussusception. Selection of these three upregulated genes was verified using quantitative reverse-transcription polymerase chain reaction (qRT-PCR). We conducted receiver operator characteristic (ROC) curve analysis to evaluate the predictive accuracy of the selected genes for pediatric intussusception. Results: We detected 732 tRFs and tRNA-derived stress-induced RNA (tiRNAs), 1705 microRNAs (miRNAs), 52 differentially expressed miRNAs, and 34 differentially expressed tRFs and tiRNAs between patients and controls. Compared with controls, we found 33 upregulated miRNAs, 24 upregulated tRFs and tiRNAs, 19 downregulated miRNAs, and 10 downregulated tRFs and tiRNAs in children with intussusception. Using qPCR, the expression trends of tRF-Leu-TAA-006, tRF-Gln-TTG-033 and tRF-Lys-TTT-028 were consistent with the sequencing results. AUCs of tRF-Leu-TAA-006, tRF-Gln-TTG-033 and tRF-Lys-TTT-028 were 0.984, 0.970 and 0.837, respectively. Conclusions: Circulating tRF-Leu-TAA-006, tRF-Gln-TTG-033 and tRF-Lys-TTT-028 expression might be a novel potential biomarker for diagnosis of pediatric intussusception.

Overcoming biological barriers by virus-like drug particles for drug delivery.

Virus-like particles (VLPs) have natural structural antigens similar to those found in viruses, making them valuable in vaccine immunization. Furthermore, VLPs have demonstrated significant potential in drug delivery, and emerged as promising vectors for transporting chemical drug, genetic drug, peptide/protein, and even nanoparticle drug. With virus-like permeability and strong retention, they can effectively target specific organs, tissues or cells, facilitating efficient intracellular drug release. Further modifications allow VLPs to transfer across various physiological barriers, thus acting the purpose of efficient drug delivery and accurate therapy. This article provides an overview of VLPs, covering their structural classifications, deliverable drugs, potential physiological barriers in drug delivery, strategies for overcoming these barriers, and future prospects.

Sub-5-Minute Ultrafast PCR using Digital Microfluidics.

The development of a rapid and reliable polymerase chain reaction (PCR) method for point-of-care (POC) diagnosis is crucial for the timely identification of pathogens. Microfluidics, which involves the manipulation of small volumes of fluidic samples, has been shown to be an ideal approach for POC analysis. Among the various microfluidic platforms available, digital microfluidics (DMF) offers high degree of configurability in manipulating µL/nL-scale liquid and achieving automation. However, the successful implementation of ultrafast PCR on DMF platforms presents challenges due to inherent system instability. In this study, we developed a robust and ultrafast PCR in 3.7-5 min with a detection sensitivity comparable to conventional PCR. Specifically, the implementation of the pincer heating scheme homogenises the temperature within a drop. The utilization of a µm-scale porous hydrophobic membrane suppresses the formation of bubbles under high temperatures. The design of a groove around the high-temperature zone effectively mitigates the temperature interference. The integration of a soluble sensor into the droplets provides an accurate and instant in-drop temperature sensing. We envision that the fast, robust, sensitive, and automatic DMF system will empower the POC testing for infectious diseases.

Clinical presentations and outcomes of pancreaticobiliary maljunction in different pediatric age groups.

BACKGROUND: Pancreaticobiliary maljunction (PBM) is a congenital defect, with risk of developing various pancreaticobiliary and hepatic complications. The presentations of PBM in children and adults are believed to be different, but studies on PBM children of different age groups are limited. This study was to evaluate clinicopathologic characteristics and outcomes in PBM children of different ages. METHODS: A total of 166 pediatric patients with PBM were reviewed retrospectively. Clinicopathological, imaging, laboratory, surgical, and follow-up data were collected and analyzed. The patients were divided into three age groups, namely, group A (< 1 year, n = 31), group B (1-3 years, n = 63), and group C (> 3 years, n = 72). RESULTS: The major clinical manifestation was jaundice in group A and abdominal pain and vomiting in groups B and C. Acute pancreatitis was more often seen in group C than group A. The length of common channel was significantly longer in group C than group A, while the maximum diameter of common bile duct in group C was smaller than that in group A. Cholangitis and cholecystitis were more commonly performed in groups B and C, while hepatic fibrosis in group A. Whether preoperatively or postoperatively, group C was more likely to have elevated serum amylase, while groups A and B were more likely to present with abnormal liver function indicators, including the increase of aspartate transaminase, alanine transaminase, and gamma-glutamyl transpeptidase. CONCLUSION: Presentation of PBM varies among different pediatric age groups, thus suggesting that targeted management should be carried out according to these differences.

Bioorthogonal Engineered Virus-Like Nanoparticles for Efficient Gene Therapy.

Gene therapy offers potentially transformative strategies for major human diseases. However, one of the key challenges in gene therapy is developing an effective strategy that could deliver genes into the specific tissue. Here, we report a novel virus-like nanoparticle, the bioorthgonal engineered virus-like recombinant biosome (reBiosome), for efficient gene therapies of cancer and inflammatory diseases. The mutant virus-like biosome (mBiosome) is first prepared by site-specific codon mutation for displaying 4-azido-L-phenylalanine on vesicular stomatitis virus glycoprotein of eBiosome at a rational site, and the reBiosome is then prepared by clicking weak acid-responsive hydrophilic polymer onto the mBiosome via bioorthogonal chemistry. The results show that the reBiosome exhibits reduced virus-like immunogenicity, prolonged blood circulation time and enhanced gene delivery efficiency to weakly acidic foci (like tumor and arthritic tissue). Furthermore, reBiosome demonstrates robust therapeutic efficacy in breast cancer and arthritis by delivering gene editing and silencing systems, respectively. In conclusion, this study develops a universal, safe and efficient platform for gene therapies for cancer and inflammatory diseases.

Exploration of the etiology of single small subcortical infarctions using high-resolution vessel wall MRI.

Objective: We aimed to explore imaging indicators for diagnosing the etiology of single small subcortical infarctions (SSI) using high-resolution vessel wall imaging (HR-VWI). Methods: Patients with acute isolated subcortical cerebral infarction were prospectively enrolled and classified as having large artery atherosclerosis (LAA), stroke of undetermined etiology (SUD), or small artery disease (SAD). The infarct information, the cerebral small vessel disease (CSVD) score, morphological characteristics of the lenticulostriate arteries (LSAs), and plaque characteristics were compared between the three groups. Results: Seventy seven patients were enrolled (30 LAA, 28 SUD, and 19 SAD). The total CSVD score of the LAA (P = 0.001) and SUD groups (P = 0.017) was significantly lower than that of the SAD group. The number and total length of LSA branches in the LAA and SUD groups were shorter than in the SAD group. Moreover, the total length laterality index (LI) of the LSAs in the LAA and SUD groups was greater than in the SAD group. The total CSVD score and LI of total length were independent predictors for the SUD and LAA groups. The remodeling index of the SUD group was significantly higher than that of the LAA group (P = 0.002); positive remodeling was dominant in the SUD group (60.7%), whereas remodeling in the LAA group was primarily non-positive (83.3%). Conclusions: SSI with and without plaques on the carrier artery may have different modes of pathogenesis. Patients with plaques may also have a coexisting mechanism of atherosclerosis.

Quantum-Squeezing-Induced Point-Gap Topology and Skin Effect.

We theoretically predict the squeezing-induced point-gap topology together with a symmetry-protected Z_{2} "skin effect" in a one-dimensional (1D) quadratic-bosonic system. Protected by a time-reversal symmetry, such a topology is associated with a novel Z_{2} invariant (similar to quantum spin-Hall insulators), which is fully capable of characterizing the occurrence of the Z_{2} skin effect. Focusing on zero energy, the parameter regime of this skin effect in the phase diagram just corresponds to a "real- and point-gap coexisting topological phase." Moreover, this phase associated with the symmetry-protected Z_{2} skin effect is experimentally observable by detecting the steady-state power spectral density. Our Letter is of fundamental interest in enriching non-Bloch topological physics by introducing quantum squeezing and has potential applications for the engineering of symmetry-protected sensors based on the Z_{2} skin effect.

Identification and validation of radiomic features from computed tomography for preoperative classification of neuroblastic tumors in children.

BACKGROUND: To identify radiomic features that can predict the pathological type of neuroblastic tumor in children. METHODS: Data on neuroblastic tumors in 104 children were retrospectively analyzed. There were 14 cases of ganglioneuroma, 24 cases of ganglioneuroblastoma, and 65 cases of neuroblastoma. Stratified sampling was used to randomly allocate the cases into the training and validation sets in a ratio of 3:1. The maximum relevance-minimum redundancy algorithm was used to identify the top 10 of two clinical features and 851 radiomic features in portal venous-phase contrast-enhanced computed tomography images. Least absolute shrinkage and selection operator regression was used to classify tumors in two binary steps: first as ganglioneuroma compared to the other two types, then as ganglioneuroblastoma compared to neuroblastoma. RESULTS: Based on 10 clinical-radiomic features, the classifier identified ganglioneuroma compared to the other two tumor types in the validation dataset with sensitivity of 100.0%, specificity of 81.8%, and an area under the receiver operating characteristic curve (AUC) of 0.875. The classifier identified ganglioneuroblastoma versus neuroblastoma with a sensitivity of 83.3%, a specificity of 87.5%, and an AUC of 0.854. The overall accuracy of the classifier across all three types of tumors was 80.8%. CONCLUSION: Radiomic features can help predict the pathological type of neuroblastic tumors in children.

Using machine learning models to predict the surgical risk of children with pancreaticobiliary maljunction and biliary dilatation.

PURPOSE: To develop machine learning (ML) models to predict the surgical risk of children with pancreaticobiliary maljunction (PBM) and biliary dilatation. METHODS: The subjects of this study were 157 pediatric patients who underwent surgery for PBM with biliary dilatation between January, 2015 and August, 2022. Using preoperative data, four ML models were developed, including logistic regression (LR), random forest (RF), support vector machine classifier (SVC), and extreme gradient boosting (XGBoost). The performance of each model was assessed via the area under the receiver operator characteristic curve (AUC). Model interpretations were generated by Shapley Additive Explanations. A nomogram was used to validate the best-performing model. RESULTS: Sixty-eight patients (43.3%) were classified as the high-risk surgery group. The XGBoost model (AUC = 0.822) outperformed the LR (AUC = 0.798), RF (AUC = 0.802) and SVC (AUC = 0.804) models. In all four models, enhancement of the choledochal cystic wall and an abnormal position of the right hepatic artery were the two most important features. Moreover, the diameter of the choledochal cyst, bile duct variation, and serum amylase were selected as key predictive factors by all four models. CONCLUSIONS: Using preoperative data, the ML models, especially XGBoost, have the potential to predict the surgical risk of children with PBM and biliary dilatation. The nomogram may provide surgeons early warning to avoid intraoperative iatrogenic injury.

Uncertainty-Aware Multi-Dimensional Mutual Learning for Brain and Brain Tumor Segmentation.

Existing segmentation methods for brain MRI data usually leverage 3D CNNs on 3D volumes or employ 2D CNNs on 2D image slices. We discovered that while volume-based approaches well respect spatial relationships across slices, slice-based methods typically excel at capturing fine local features. Furthermore, there is a wealth of complementary information between their segmentation predictions. Inspired by this observation, we develop an Uncertainty-aware Multi-dimensional Mutual learning framework to learn different dimensional networks simultaneously, each of which provides useful soft labels as supervision to the others, thus effectively improving the generalization ability. Specifically, our framework builds upon a 2D-CNN, a 2.5D-CNN, and a 3D-CNN, while an uncertainty gating mechanism is leveraged to facilitate the selection of qualified soft labels, so as to ensure the reliability of shared information. The proposed method is a general framework and can be applied to varying backbones. The experimental results on three datasets demonstrate that our method can significantly enhance the performance of the backbone network by notable margins, achieving a Dice metric improvement of 2.8% on MeniSeg, 1.4% on IBSR, and 1.3% on BraTS2020.