Safety valves are critical safety devices in pressure-bearing equipment used in the energy sector. Due to long-term erosion and wear by the medium, the sealing surfaces of their valve seats are prone to damage, which leads to internal leakage. This not only causes ineffective loss of energy medium and exacerbates the pressure on energy conservation and consumption reduction, but also poses a significant safety hazard, threatening the stable operation of the energy system. Acoustic emission signal technology, as a mainstream non-destructive testing method, has been widely used in the detection of internal leakage in safety valves. However, it lacks theoretical support for the acoustic characteristics of internal leakage aerodynamic noise, which limits the accuracy and reliability of the detection. This paper studies the internal leakage aerodynamic noise of a spring-loaded full-lift safety valve. Large eddy simulation (LES) is used to perform numerical calculations of transient internal leakage flow to capture velocity pulsations in the fluid domain. The Lighthill tensor is solved using a energy-conserving mapping method as the excitation source, and the acoustic solution is completed by combining the Ffowcs Williams-Hawkings (FW-H) equations to obtain the sound pressure level curve and contour map of internal leakage aerodynamic noise. The study found that the sound pressure pulsation of internal leakage is dominated by quadrupoles, and the main peak frequency of internal leakage is negatively correlated with the total sound pressure level. The relative error between the numerical simulation and experimental results is ≤ 5%, indicating good accuracy. This research fills a theoretical gap in the acoustic characteristics of internal leakage, provides theoretical guidance for the acoustic detection of internal leakage, can improve the accuracy of leakage detection, achieve early warning, reduce energy waste, help save energy and reduce consumption, and has important engineering application value for ensuring the safe and stable operation of energy systems.
Retrospective cohort study. To identify risk factors for revision surgery due to persistent cerebrospinal fluid (CSF) leakage after an intraoperatively recognized and repaired incidental durotomy (ID) in lumbar spine surgery. ID is a common complication in spine surgery, often leading to CSF leakage, prolonged hospitalization, and increased morbidity. While risk factors for ID are well described, limited data exist on predictors of persistent CSF leakage requiring revision surgery. A total of 323 patients who sustained an intraoperatively recognized and repaired ID during lumbar spine surgery were retrospectively analyzed. Patients with traumatic dural tears or bacterial infections were excluded. Demographic data, comorbidities, surgical details, and complications were recorded. Univariate and multivariate logistic regression analyses identified independent risk factors for revision surgery. Fifteen patients (4.64%) required revision surgery for persistent CSF leakage. Univariate analysis identified type 2 diabetes, chronic obstructive pulmonary disease, chronic kidney disease, and metastatic tumor history as risk factors. Multivariate analysis confirmed type 2 diabetes (aOR: 3.732; 95% CI: 1.093-12.739; P = 0.036), metastatic tumor history (aOR: 15.758; 95% CI: 2.339-106.160; P = 0.005), and male sex (aOR: 4.018; 95% CI: 1.022-15.794; P = 0.046) as independent risk factors. Revision patients had significantly longer hospital stays (16 vs. 8 d, P = 0.014) and higher rates of wound infection and ischemic stroke. Type 2 diabetes, metastatic tumor history, and male sex are independent risk factors for revision surgery due to persistent CSF leakage after ID repair. These findings highlight the need for increased vigilance and potentially modified dural repair strategies in high-risk patients.
Cement leakage is a common and occasionally serious complication of percutaneous vertebroplasty (PVP). We compared cement leakage (primary outcome) and post-operative pain (VAS) and disability (ODI) (secondary outcomes) between trial-defined lower-volume and higher-volume cement injection in osteoporotic vertebral compression fractures. We searched PubMed, Embase, Scopus, the Cochrane Library, CNKI and Wanfang from January 2018 to 31 December 2025 (last search date), and included randomized controlled trials (RCTs). Risk of bias was assessed with RoB 2. "Lower-volume" denotes each trial's smaller-volume arm. Reported lower-volume arms ranged from 1.5 to 2.5 mL to < 6 mL, whereas higher-volume arms ranged from 2.6 to 3.5 mL to 7.0 mL or ≥ 6 mL, depending on the trial definition. Risk ratios (RR) for leakage and mean differences (MD) for VAS/ODI were pooled with random-effects models; an exploratory subgroup (lower-arm ≤ vs. > ~ 3.5 mL) was examined, and sensitivity analyses included leave-one-out analysis, exclusion of the single cervical trial, and exclusion of trials with ambiguous volume labels. Differences in VAS/ODI were interpreted against established minimal clinically important difference (MCID) thresholds. Sixteen RCTs were included. Lower-volume injection was associated with a markedly lower risk of cement leakage (RR 0.30, 95% CI 0.22-0.40; I² = 0%), consistent across subgroups (test for interaction P > 0.30) and stable in leave-one-out analysis (RR 0.27-0.31) and on exclusion of the single cervical trial. Post-operative VAS differed negligibly between arms (MD 0.01 points, 95% CI -0.11 to 0.12), as did ODI among the 12 trials that actually measured it (MD -0.04 points, 95% CI -0.52 to 0.44); both were well below accepted MCID thresholds. Four trials did not measure ODI and were excluded from the ODI analysis. Trial-defined lower-volume cement injection was associated with a substantially reduced risk of cement leakage, while differences in post-operative pain and disability were negligible and below accepted MCID thresholds. Because no non-inferiority margin was pre-specified, these findings should not be read as statistical non-inferiority. Given the wide variation in volume definitions and procedural detail across trials, the evidence supports an individualized lower-volume strategy rather than a single universally optimal volume.
Bile leakage remains a clinically relevant complication after hepatectomy and contributes to morbidity, prolonged drainage, extended hospital stay, and the need for reintervention. Intraoperative indocyanine green (ICG) fluorescence imaging can be used to visualize bile leaks from the transection plane or biliary stump, enabling targeted repair. However, evidence for the effect of this technique on clinically relevant bile leakage is heterogeneous and has not been systematically synthesized. This protocol describes a systematic review and meta-analysis of randomised controlled trials and comparative nonrandomised studies evaluating indocyanine green fluorescence imaging-guided intraoperative bile leak detection during hepatectomy. MEDLINE, Embase, Cochrane CENTRAL, and the Web of Science Core Collection will be searched from inception, along with trial registries and citation tracking. The primary outcome is clinically relevant postoperative bile leakage, defined as International Study Group of Liver Surgery (ISGLS) grade B or C. Secondary outcomes include any bile leakage, bile leak-related interventions, major postoperative complications, length of postoperative hospital stay, and mortality. Randomised and nonrandomised studies will be synthesized separately. A meta-analysis will be performed when the studies are sufficiently comparable; otherwise, the findings will be summarized narratively. Planned analyses include random-effects models, subgroup analyses stratified according to the route of indocyanine green administration, sensitivity analyses, and an assessment of the certainty of evidence using the Grading of Recommendations Assessment, Development and Evaluation approach. This protocol is registered in PROSPERO (CRD420261291065). This manuscript describes the planned methods only; study selection, data extraction, and evidence synthesis results will be reported in the completed systematic review. This study will use data from published studies and does not require ethics approval. The findings will be disseminated through peer-reviewed publications and conference presentations.
Public leaderboards such as the Therapeutics Data Commons (TDC) ADMET benchmark are widely treated as a ranking of state-of-the-art models. However, a high leaderboard position is only meaningful if the corresponding model can actually be reproduced and deployed by an independent researcher. In this work, we audit whether the top-ranked TDC ADMET models meet that bar. We assessed the top-ranked models of all 22 TDC ADMET leaderboards from the perspective of an end user with access only to the publicly released artifacts of each model─its publication, code repository, and installation instructions. For every end point, the top three models were screened with a unified protocol including an execution environment reproducibility check, a data-leakage assessment, verification of the hyperparameter-optimization procedure, and a reevaluation against the current leaderboard. Only three models (CaliciBoost, MapLight, and MapLight + GNN) passed all stages and reproduced their reported performance. The remaining models failed because of unavailable code, nonreproducible environments, runtime incompatibilities, or methodological flaws. We traced direct or indirect data leakage in the MiniMol, GradientBoost, and XGBoost models, and used deliberately overfitted variants of our own Mol2Vec-based models to show that tuning on the public test set─whether accidental or intentional─can substantially inflate both metrics and leaderboard rank. These results indicate that current TDC leaderboard positions cannot be read as a direct measure of model quality and practical applicability and emphasize the urgent need for better public ADMET benchmarks based on the hidden test sets, strict data set versioning and model submission with standardized inference environments.
Referring Expression Segmentation (RES) requires models not only to locate objects specified by referring expressions accurately but also to predict complete and precise masks. Existing methods primarily focus on complex multimodal alignment for object grounding, often neglecting mask quality, which results in incomplete foreground regions and imprecise boundaries. To address these challenges, we propose DiffRES, a mask-generating framework based on Stable Diffusion (SD), designed to tackle the RES problem with a focus on achieving high-quality masks. DiffRES effectively mitigates the information leakage issue prevalent in existing generative dense prediction diffusion models, which allows the model to infer the target's position directly from noisy masks during training without understanding the text condition, leading to severe overfitting. Specifically, DiffRES directly guides SD with visual and linguistic information to generate target binary masks, fundamentally bypassing the information leakage issue. This approach enables efficient knowledge transfer from SD to the RES task, resulting in precisely localized binary masks with sharp and precise boundaries. Extensive experiments show that DiffRES surpasses current state-of-the-art traditional methods on boundary precision (APb) which is sensitive to mask quality, while also significantly outperforming all existing SD-based RES models across all metrics. Our code is publicly available at https://github.com/charon517-517/DiffRES.
The role of routine diverting ileostomy after restorative rectal cancer surgery remains controversial. We evaluated a selective non-diverting strategy implemented together with technical modifications of the anastomosis and an intensified postoperative surveillance protocol. This retrospective single-center study included 300 patients who underwent restorative surgery for rectal cancer between 2017 and 2026. Patients were categorized according to reconstruction technique: end-to-end stapled colorectal anastomosis, side-to-end stapled colorectal anastomosis, or delayed coloanal anastomosis using the Turnbull-Cutait technique. The primary endpoint was clinically relevant anastomotic leakage, defined according to the International Study Group of Rectal Cancer classification as Grade B or C leakage. Fisher exact, chi-square, Mann-Whitney U, and Kruskal-Wallis tests were used as appropriate; 95% confidence intervals (CIs) were calculated for leakage rates. The cohort included 71 end-to-end, 171 side-to-end, and 58 delayed coloanal reconstructions. Clinically relevant leakage occurred in 8/71 patients after end-to-end anastomosis (11.3%, 95% CI 5.0-21.0), 4/171 after side-to-end anastomosis (2.3%, 95% CI 0.6-5.9), and 2/58 after delayed coloanal anastomosis (3.4%, 95% CI 0.4-11.9; p = 0.010). Among non-diverted stapled colorectal anastomoses, clinically relevant leakage was lower after side-to-end reconstruction than after end-to-end reconstruction (1/163 [0.6%] vs. 3/45 [6.7%], p = 0.032). In selected patients, a non-routine diversion strategy combined with optimized anastomotic configuration, routine splenic flexure mobilization when required, delayed coloanal reconstruction for ultra-low tumors, and intensive postoperative surveillance appears feasible. Because several technical and organizational changes were introduced simultaneously, these findings should not be interpreted as proving the independent safety of omitting diversion.
Central Serous Chorioretinopathy(CSC) is a chorioretinal disorder, predominantly affects young to middle-aged adults, resulting serious vision disorder. This study aimed to develop a Bayesian network model to predict the key factors influencing the early therapeutic efficacy of 577 nm-SML in patients with CSC. A total of 159 patients (159 eyes) diagnosed with CSC and treated with 577 nm-SML at the First Affiliated Hospital of Guangxi Medical University from January 2019 to November 2023 were retrospectively analyzed. Baseline data including age, sex, eye side, disease course, and past medical history were collected. Ophthalmic imaging detects central macular thickness (CMT), macular foveal volume (MFV within 1mm, 3mm, 6mm diameter), height and area of subretinal fluid (SRF), structural changes in the outer retinal layers (ORL), type and area of leakage lesions, etc. Influential variables significantly associated with 577nm-SML efficacy were screened using LASSO regression, then construct a Bayesian network model to predict factors that significantly affect the therapeutic effect. LASSO regression identified 19 significant variables related to treatment outcomes from the 40 possible risk factors included, including disease duration, sex, eye Side, smoking, hormone, macular foveal volumes (3 mm and 6 mm diameters), and the height and area of SRF, ORL integrity, typical PED, location of PED, location of RPE bulging, heterogeneity of NPL, HF of ORL, HF of choroid, leakage type, leakage location, leakage correlate with OCT. The Bayesian network presents complex interactions among these factors, shows that patients with smaller macular foveal volumes (within 3 mm diameter), shorter disease duration, and focal leakage exhibited superior responses to 577nm-SML treatment. The therapeutic response to 577nm-SML in CSC is influenced by multifactorial dynamics. Bayesian network can well present the complex network relationship between the therapeutic effect of 577nm-SML and related influencing factors, and identify potential risk factors that affect early efficacy.
Computational models that predict cancer drug response from genomic features are central to biomarker discovery, yet a recent audit found data leakage in 72% of 32 published methods, and complex models offer little interpretability while only modestly exceeding simple baselines under honest evaluation. Tissue lineage is a largely untapped source of legitimate inductive bias, but existing tissue-aware methods neither separate pan-cancer from lineage-specific signal nor report leakage-free performance. We introduce the Data Shared Elastic Net (DSEN), a tissue-aware regression that decomposes each drug's model into a shared coefficient block common to all lineages and tissue-specific deviation blocks. Under leakage-free cross-validation across 265 drugs, 1,462 cell lines and 31 tissue lineages, DSEN improved mean squared error over a standard elastic net for 92.5% of drugs (mean 4.95%) while selecting 58% fewer stable shared features. Shared coefficients generalized to held-out tissues (59% tissue-level win rate) and recurrently recovered transferable pathway modules (p53, MAPK), whereas tissue blocks captured lineage markers such as the skin MITF / S100B program. The closest tissue-aware comparator, TG-LASSO, performed worse than the tissue-agnostic baseline (-13.8% mean MSE). Ablation shows tissue-aware modeling helps most when features are scarce, with no single modality dominating. https://github.com/AsiaeeLab/tissue-aware-drug-response . amir.asiaeetaheri@vumc.org.
Esophagectomy remains a highly invasive procedure associated with substantial postoperative morbidity. Pulmonary complications, anastomotic leakage, and in-hospital mortality are of particular concern. Perioperative corticosteroids are often administered to attenuate excessive inflammatory responses; however, the clinical impact in esophageal cancer surgery remains inconclusive, with studies yielding inconsistent findings and lacking integrated long-term evaluation. This meta-analysis aimed to comprehensively assess the effects of perioperative corticosteroid administration on both short-term complications and long-term survival in patients undergoing esophagectomy. A systematic literature search identified randomized controlled trials and observational studies published to October 2025 that examined corticosteroid use during esophagectomy. Primary outcomes comprised postoperative pneumonia, anastomotic leakage, and in-hospital mortality. Secondary outcomes comprised respiratory failure, acute respiratory distress syndrome, wound infection, hepatic and renal dysfunction, cardiac complications, and overall survival. Pooled odds ratios (ORs) and hazard ratios were calculated using random-effects models. Nineteen studies involving 38 316 patients were included. Perioperative corticosteroid use was associated with a significantly lower incidence of pneumonia (OR = 0.71), in-hospital mortality (OR = 0.79), and respiratory failure (OR = 0.30) compared with no corticosteroids. No increase in anastomotic leakage or wound infection rates were observed with versus without perioperative corticosteroids; long-term survival was unaffected (hazard ratio: 0.85), and no excess risk of other complications was detected. To our knowledge, this is the first meta-analysis integrating both short- and long-term outcomes, suggesting that perioperative corticosteroids may reduce major postoperative complications without evidence of adverse long-term oncologic outcomes; however, these findings were largely driven by observational studies.
To clarify the recharge sources, water cycle evolution mechanisms, and water-rock interaction patterns of typical karst spring basins in the northern Taihang Mountains, this study investigated the Shuimocao karst spring group through a 1:50,000 hydrogeological survey and continuous dynamic monitoring of spring and river discharge. A total of 38 hydrochemical samples and 26 hydrogen and oxygen isotope samples were collected and tested. Integrating high-density resistivity profiling, hydrochemical identification, isotope tracing, and hydrological dynamic analysis, this study systematically explored the basin tectonic evolution, karst development features, hydrochemical signatures, and groundwater water cycle mechanisms, while elucidating the driving factors responsible for spring drying up. The results indicate that Meso-Cenozoic tectonic movements established the fundamental tectonic framework of the study basin. Highly soluble limestone of the Ordovician Majiagou Formation provides essential material conditions for karst development, whereas concealed faults and epikarst zones constitute dominant migration pathways for surface water and groundwater. Major hydrochemical ions are primarily derived from carbonate and silicate rock weathering, accompanied by staged cation exchange, while anthropogenic activities significantly disturb water quality in the spring discharge zone. Quantitative end-member mixing analysis (EMMA) identifies three groundwater recharge pathways: leakage from the Tongtian River (51.93%) serves as the primary recharge source, followed by Sanhui River leakage (28.79%) and lateral groundwater runoff from the Yebei area (19.29%). Precipitation only acts as an indirect, time-lagged driving factor rather than a direct water source. Overall, the spring system is predominantly recharged by river infiltration and secondarily by lateral groundwater inflow, forming a tectonically controlled water cycle pattern: "surface water infiltration - karst conduit migration - flow regulation by faults and aquicludes - concentrated spring discharge". The extreme spring drying event in 2020 resulted from the superposition of extreme drought and persistent groundwater overexploitation, and groundwater flow field alteration induced by water level decline was the direct cause of well water turbidity. This study reveals the surface water-groundwater coupling mechanism of typical northern karst basins, providing a representative case and scientific reference for karst groundwater exploitation, spring ecological conservation, and analogous hydrogeological studies in the Taihang Mountains.
Tubeless strategy of thoracoscopic surgery has become a new topic in recent years. This study aimed to investigate the feasibility of no chest tube drainage compared with chest tube drainage in patients who underwent uniportal thoracoscopic sublobar resection. Patients who underwent uniportal thoracoscopic sublobar resection with or without chest tube drainage were included in this retrospective cohort study. They were required to have a negative intraoperative air leakage test. Data regarding perioperative outcomes, postoperative complications, postoperative pain visual analogue scale (VAS) score at rest, and patients' satisfaction were collected. A total of 60 patients without chest tube drainage served as the tubeless group, while another 60 age/sex-matched patients with chest tube drainage served as the chest tube group. Pneumothorax occurred in only one (1.7%) patient in the tubeless group, which was not significantly different from 0 (0.0%) patient in the chest tube group (P = 0.500). This patient was further readmitted and underwent conversion to chest tube drainage. No patient experienced other postoperative complications such as pleural effusion and pulmonary infection in either group. The operational time (P < 0.001), perioperative blood loss (P = 0.007), total hospital stay (P = 0.001), and postoperative hospital stay (P < 0.001) were lower in the tubeless group than in the chest tube group. The pain VAS score was decreased in the tubeless group compared with the chest tube group on postoperative day (POD) 1 (P = 0.003) and 2 (P = 0.020), but not changed on POD 3 (P = 0.200). Moreover, patients' satisfaction scale (P = 0.410) and overall satisfaction rate (P = 0.375) were not different between the two groups. After adjustment using multivariable regressions, tubeless versus chest tube was independently correlated with a lower operational time (P < 0.001), total hospital stay (P = 0.003), postoperative hospital stay (P < 0.001), and pain VAS score on POD1 (P = 0.009) and POD2 (P = 0.042) but was not independently associated with perioperative blood loss (P = 0.110), pain VAS score on POD3 (P = 0.348), pneumothorax presence (P = 0.997), or overall satisfaction (P = 0.553). Uniportal thoracoscopic sublobar resection without chest tube drainage may be feasible in the selected, low-risk patients without intraoperative air leakage. However, further prospective studies with standardized outcome capture are still needed for validations.
Background Incidental dural tears are a recognized complication of lumbar spine surgery and can lead to significant postoperative morbidity if not properly managed. Standard repair techniques often rely on specialized grafts and sealants that may be costly or unavailable in low-resource settings. This study reports our experience with the use of cyanoacrylate adhesive (superglue) for intraoperative repair of dural tears in a resource-limited environment. Methods We retrospectively reviewed patients who sustained incidental dural tears during lumbar spine surgery at a tertiary spine centre in a resource-limited setting between January 2022 and June 2025. All tears were identified intraoperatively and repaired immediately. In six cases, cyanoacrylate adhesive was used as part of the dural repair strategy. In two patients, lumbar fascia grafts were used to cover inaccessible tears and reinforced with Surgicel™ (Ethicon, Inc., Raritan, New Jersey) before application of cyanoacrylate adhesive. In four patients, autologous fat grafts and Surgicel™ were similarly used prior to cyanoacrylate application. Postoperative outcomes, including cerebrospinal fluid (CSF) leakage, wound infection, pseudomeningocele formation, and need for reoperation, were reviewed. Results Among 68 patients who underwent lumbar spine surgery, nine (13.2%) sustained intraoperative dural tears, and six of these were managed using cyanoacrylate-assisted repair. Watertight closure was achieved intraoperatively in all cases. One patient developed a mild superficial wound infection that resolved with antibiotics, while another developed a pseudomeningocele following revision surgery that was managed conservatively. There were no cases of persistent CSF leakage, meningitis, or reoperation. Conclusion Cyanoacrylate-assisted repair appeared to provide a feasible and low-cost adjunct for dural repair in this limited case series. When used in conjunction with autologous grafts and Surgicel™, satisfactory short-term outcomes were observed without major complications. Further prospective studies are needed to better establish long-term safety and effectiveness.
Age-related macular degeneration (AMD) is a leading cause of vision loss, with its neovascular form (nAMD) primarily treated using anti-VEGF agents; however, therapeutic resistance and nonresponse remain major clinical challenges. Tanshinone IIA (TIIA), a multi-target bioactive compound derived from Salvia miltiorrhiza, has shown potential in retinal disease treatment. In this study, we investigated the therapeutic effects and underlying mechanisms of TIIA on choroidal neovascularization (CNV) using Vldlr knockout (Vldlr-/-) mice as an nAMD model. TIIA was administered intraperitoneally for 8 weeks, and CNV progression and vascular leakage were evaluated by OCT and FFA, while outer blood-retinal barrier (oBRB) integrity was assessed by immunofluorescence staining. Proteomics analysis combined with western blotting was used to explore the molecular mechanisms. Our results showed that TIIA significantly reduced CNV area and leakage, and restored oBRB integrity by upregulating tight junction proteins ZO-1 and Occludin in the RPE/choroid complex. Mechanistically, TIIA inhibited angiogenesis via suppression of the PLCγ/ERK1/2 signaling pathway. In addition, proteomics analysis revealed enhanced cholesterol efflux, intermediate filament reorganization, and decreased autophagy-related proteins across the retina, RPE/choroid complex, and serum. Collectively, these findings demonstrate that TIIA alleviates nAMD pathology through multi-target mechanisms, including inhibition of angiogenesis, restoration of barrier function, metabolic reprogramming, and modulation of autophagy, highlighting its potential as an alternative therapeutic strategy for nAMD.
Anastomotic leakage remains challenging in left-sided colorectal surgery. We previously proposed a 5-grade visual grading system based on marginal vessel bleeding, indicating that good (grades A/B) to moderate (grade C) perfusion generally ensures safe anastomosis. However, borderline perfusion (grade C) carried additional risk, particularly among older patients or those with comorbidities. We refined this system to determine whether selective resection of grade C bowel could improve perfusion and guide ostomy diversion decisions. Eighty patients with left-sided colorectal cancer underwent curative-intent surgery. Intraoperatively, both marginal vessel and mucosal bleeding were assessed. Initial perfusion grade-guided decision-making: grades A/B proceeded to anastomosis, whereas for grade C, additional resection was performed to upgrade perfusion. Ostomy diversion was performed for persistent grade C or prophylactically in high-risk patients. Among 80 patients, 67 (83.8%) had grades A/B and 13 (16.3%) were initial grade C. Additional resection (mean, 4.5 cm) upgraded perfusion to final grade B in 9 patients (69.2%), enabling primary anastomosis. Overall, 12 patients (15.0%) received ostomy: 4 for persistent grade C and 8 for high-risk indications despite adequate final perfusion. All 76 patients with final grade A/B achieved 0% anastomotic leakage and ischemic colitis. Grade C patients had longer operative times (108.8 minutes vs. 86.6 minutes, P=0.003) with additional procedures. Patients requiring diversion were older with more comorbidities. By actively applying the intraoperative visual grading system to upgrade borderline perfusion, we achieved low postoperative complication rates. This approach provides a simple and practical strategy to support intraoperative clinical judgment and optimize anastomotic safety.
Postpartum depression (PPD) affects nearly one in five mothers, yet many cases remain undetected. This study demonstrates the value of a leakage-resistant, rigorously designed machine learning pipeline for early prediction using psychosocial data from 1,430 women during pregnancy. Concept-level feature harmonization addressed data artifacts, while repeated group-aware cross-validation, calibration, and recall-oriented thresholding ensured robust evaluation. Class imbalance was handled using weighting and SMOTE. Models showed strong performance (ROC-AUC: 0.762-0.801; PR-AUC: 0.400-0.490), with logistic regression and ensemble methods performing best. High recall (>0.90) supports screening use, though moderate precision highlights the need for clinical follow-up. SHAP analysis identified anxious attachment and coping factors as key predictors. Overall, well-designed ML pipelines can enable early, reliable PPD risk stratification while minimizing bias and leakage.
Federated Learning (FL) emerged as a privacy-preserving paradigm for collaborative training of deep learning models across institutions without sharing patient data. This approach has been applied to complex tasks such as medical image-to-image (I2I) translation, including MRI-to-synthetic CT (sCT) generation. However, existing federated I2I frameworks often assume privacy preservation as an inherent property of FL rather than a requirement to be explicitly validated, leaving their robustness to representative adversarial threat scenarios largely unexplored. In this study, we evaluated the vulnerability of a federated MRI-to-sCT translation framework (FedSynthCT-Brain) to three representative attack classes: Deep Leakage from Gradients (DLG), Federated Membership Inference Attack (FedMIA), and data poisoning. The efficacy of corresponding defense mechanisms, such as Secure Aggregation (SecAgg) and Byzantine-robust median aggregation (FedMedian), were assessed. DLG enabled only the recovery of coarse anatomical structures, with no clinically identifiable details (SSIM ≤ 0.16, PSNR ≤ 11 dB) across clients, suggesting limited vulnerability under the evaluated DLG setting. In contrast, FedMIA achieved high membership discrimination, with AUC scores between 0.92 and 0.99, revealing a critical privacy vulnerability. The introduction of SecAgg reduced AUC values to near-random levels (0.23-0.56) across all centers without impacting synthesis quality. Under high-noise poisoning, the standard federated averaging (FedAvg) aggregation rendered the federation inoperative, while FedMedian restored performance close to the no-poisoning baseline in most scenarios, with significant residual degradation in specific center configurations. At low noise levels, the advantage of FedMedian was less consistent, as low-level noise injection may be indistinguishable from natural heterogeneity across centers, potentially enabling stealthy degradation. These findings demonstrate that federated I2I translation frameworks are not inherently secure and require explicit, multi-layered evaluation. As FL is increasingly adopted in clinical workflows, our results underscore the necessity of integrating cryptographic, algorithmic, and infrastructural safeguards for secure deployment.
The interdigitated back contact (IBC) structure offers high efficiency potential for crystalline silicon solar cells, yet its low-light performance (LLP) faces ongoing debates. This study systematically investigates LLP mechanisms in back contact (BC) cells via process optimization, simulation, and machine learning (ML). We identify leakage paths caused by residual "cap-shaped" borosilicate glass in rear p-type poly-Si regions as a critical bottleneck. By optimizing laser grooving through gap adjustments and additive engineering, the LLP of tunnel oxide passivated contact (TBC) cells is elevated to match that of TOPCon cells, reducing power loss to below 5%. Using SunSolve and Quokka3 simulations, we analyze module-level double-diode parameters and device-to-module losses. ML-driven data mining reveals an inherent trade-off between LLP and power conversion efficiency (PCE). To resolve this, we develop a multi-head neural network integrated with an evolutionary algorithm for co-optimization. This yields a candidate parameter set that effectively balances PCE and LLP across both TOPCon and TBC solar cells. Our work clarifies the microscopic origins of LLP limitations and provides a practical framework for designing high-efficiency BC cells with superior low-light response, accelerating industrial application.
Post-stroke urinary incontinence is a prevalent and debilitating complication that severely impacts patient recovery and quality of life. Although pelvic floor muscle training is a recognized standard rehabilitative intervention, its clinical efficacy in stroke survivors is often suboptimal, largely because it addresses peripheral muscle function without mitigating the underlying central neurological deficits that disrupt micturition control. Emerging evidence suggests that neuromodulation via transcranial direct current stimulation can enhance neuroplasticity in cortical motor areas. This trial aims to address the current therapeutic gap by evaluating whether the synergistic combination of central neuromodulation and peripheral muscle training provides superior therapeutic benefits for post-stroke urinary incontinence recovery compared to pelvic floor muscle training alone. This randomized controlled trial will randomly assign an aggregate of 50 eligible patients into two groups. Participants allocated to the intervention group (n = 25) will undergo PFMT coupled with active, low-frequency tDCS over M1, whereas those in the control group will receive identical PFMT alongside sham tDCS. The 20 treatment sessions will be administered on a schedule of five sessions per week across four consecutive weeks for every participant. The primary outcome is leakage episodes. The secondary outcomes are defined as simplified cystometry, the International Consultation on Incontinence Questionnaire Urinary Incontinence Short Form (ICIQ-UI SF), Overactive Bladder Symptom Score (OABSS), Incontinence Quality of Life Questionnaire Score (I-QOL), surface electromyography (sEMG), Neurogenic Bowel Dysfunction Score and the number of urinations. This randomized controlled trial protocol evaluates a novel intervention combining tDCS targeting the contralesional M1 with PFMT for post-stroke urinary incontinence. To address a critical gap in stroke rehabilitation, the study designs a rigorous double-blind, sham-controlled design to explore the combined effects of central neuromodulation and peripheral muscle training. A comprehensive assessment will be used to generate robust evidence of this innovative intervention. The findings are expected to contribute to an effective neurorehabilitation strategy, potentially transforming clinical approaches to this common yet under-addressed complication and significantly enhancing the quality of life for stroke survivors. URL: https://www.chictr.org.cn; Unique identifier: ChiCTR2500099178. Registered on 2025-03-19.
Brain tumors are a leading cause of cancer-related mortality, and manual MRI screening remains time-consuming and observer-dependent. Deep learning (DL) offers automated detection, but clinical translation requires rigorous validation and interpretability. This study introduces a DL framework for brain tumor detection that addresses two major challenges in medical AI: limited dataset availability and lack of interpretability. Preliminary experiments identified InceptionV3 optimized with Nadam as the optimal architecture. To ensure robust validation, this model was retrained using patient-wise stratified fivefold cross-validation on 90% of the data incorporating augmentation and minority oversampling to prevent data leakage. This achieved an overall accuracy of 98.3 ± 0.9%. The final model was then trained on the entire development set using the optimal configuration, thereby leveraging all available labeled data to maximize learning capacity and enhance generalization. Performance evaluation was conducted on three levels: (i) a held out internal test set (10% of the data) for internal assessment, (ii) an external dataset of 3000 unseen images for independent validation, and (iii) quantitative explainable AI (XAI) analyses performed on both internal and external test datasets. The proposed model achieved perfect classification metrics on the internal test set, with 100% accuracy and minimal loss (0.01), and demonstrated strong generalizability on the external dataset with 96% accuracy and minimal loss (0.11). Quantitative XAI analysis demonstrated high faithfulness (Grad-CAM vs. occlusion sensitivity correlation exceeded 0.8), causal importance (top-10% occlusion drop 44% vs. 9% for random occlusion), and specificity to learned weights (Spearman correlation ≈ - 0.01). The proposed pipeline establishes a rigorous, transparent framework for data-limited medical imaging, demonstrating high diagnostic performance with clinically aligned explanations and providing a reliable foundation for trustworthy AI in brain tumor detection.