Category: Statistics

Surg Endosc. 2025 Aug 7. doi: 10.1007/s00464-025-12021-8. Online ahead of print.

ABSTRACT

BACKGROUND: Traditional intraoperative navigation methods show insufficient adaptability in dynamic surgical environments. The rapid development of Artificial Intelligence (AI) presents an opportunity to overcome these limitations, making the construction of real-time, adaptive intraoperative navigation systems a key research goal. This study, based on Edge-AI and multimodal AI technologies, aims to develop and evaluate a foundational system for achieving real-time, offline intraoperative navigation and warnings during minimally invasive liver surgery.

METHODS: 161 minimally invasive liver resection videos collected from the Medtechshare platform were finely annotated. The dataset was divided into training (60%), validation (20%), and independent test (20%) sets. The self-developed ONE-PEACE-SZYYv3 model was used for training and validation, followed by quantitative evaluation using accuracy, recall, precision, Intersection over Union (IoU), and Dice coefficient on the independent test set. For qualitative evaluation, an innovative Turing-like test method, the Humanlike Test, was proposed. In this test, experienced surgeons blindly evaluated and scored AI-generated warnings and manually delineated warnings created by junior surgeons. Two groups of data from different processing sources were blindly evaluated by experts, combined with Mann-Whitney U and TOST (Δ = ± 1.5 points) tests to assess whether they were equivalent within an acceptable range. This ultimately tested whether the current AI possesses judgment capabilities plausible enough to be comparable to human levels during surgery.

RESULTS: In quantitative evaluation on the independent test set, under mAP50, the model’s mean average precision reached as high as 99.5%, and under the stricter mAP50-95, it achieved 89.9% precision. IoU and Dice were 0.63 and 0.79, respectively. In qualitative evaluation, the average score of the “AI guided results, Experimental group” was slightly lower than the “Manually delineated results, Control group,” but the difference was not statistically significant (p > 0.05). Equivalence testing confirmed that the scores were statistically equivalent within a pre-defined narrow margin.

CONCLUSION: The proposed model exhibits high precision and real-time capability on a curated, retrospective dataset. It can accurately and effectively provide multimodal guidance (image-text-sound) and warnings in complex anatomical structures with multiple types of interference elements during surgery. The innovative Humanlike Test shows its intraoperative judgment capability can, to a certain extent, produce outputs that are indistinguishable from those of human surgeons. While this study demonstrates foundational feasibility, prospective clinical trials are required to validate its clinical utility and impact on surgical outcomes. Such technology holds promise for elevating the surgical field to a new paradigm of digital intelligence.

PMID:40775470 | DOI:10.1007/s00464-025-12021-8

Br J Cancer. 2025 Aug 7. doi: 10.1038/s41416-025-03141-y. Online ahead of print.

ABSTRACT

BACKGROUND: Integrating genome-wide association study (GWAS) and transcriptomic datasets can identify mediators for genetic risk of cancer. Traditional methods often are insufficient as they rely on total gene expression measures and overlook alternative splicing, which generates different transcript-isoforms with potentially distinct effects.

METHODS: We integrate multi-tissue isoform expression data from the Genotype Tissue-Expression Project with GWAS summary statistics (all N > ~20,000 cases) to identify isoform- and gene-level associations with six cancers (breast, endometrial, colorectal, lung, ovarian, prostate) and six related cancer subtype classifications (N = 12 total).

RESULTS: Directly modeling isoforms using transcriptome-wide association studies (isoTWAS) significantly improves discovery of genetic associations compared to gene-level approaches, identifying 164% more significant associations (6163 vs. 2336) with isoTWAS-prioritized genes enriched 4-fold for evolutionarily-constrained genes. isoTWAS tags transcriptomic associations at 52% more independent GWAS loci across the six cancers. Isoform expression mediates an estimated 63% greater proportion of cancer risk SNP heritability compared to gene expression. We highlight several isoTWAS associations that demonstrate GWAS colocalization at the isoform level but not at the gene level, including CLPTM1L (lung cancer), LAMC1 (colorectal), and BABAM1 (breast).

CONCLUSION: These results underscore the importance of modeling isoforms to maximize discovery of genetic risk mechanisms for cancers.

PMID:40775447 | DOI:10.1038/s41416-025-03141-y

Br J Cancer. 2025 Aug 7. doi: 10.1038/s41416-025-03110-5. Online ahead of print.

ABSTRACT

BACKGROUND: One-third of all lung cancer cases globally are reported in China. This study evaluated the symptom management efficacy of an electronic patient-reported outcomes (ePRO)-based intervention for postoperative symptoms like pain and psychological distress after lung cancer surgery.

METHODS: We included lung cancer surgery patients (April 2022-October 2023; age, 18-75 years) with ECOG scores of 0-2 and expected survival of >6 months and randomized them into control and intervention groups. The latter completed MDASI-LC and QLQ-C30 questionnaires, wherein high symptom scores prompted treatment recommendations; the former received routine care. Changes in symptom scores, daily function, and quality of life were evaluated over 12 weeks and 1 year through surveys and interviews for ePRO-based symptom management efficacy assessments.

RESULTS: Herein, 355 participants comprised intervention (n = 182) and control groups (n = 173). At 12 weeks, the former had significantly lower symptoms threshold [0 (0-1) vs. 1 (0-3)], lower symptom scores [adjusted mean difference, -0.527 (95% CI: -0.788 to -0.266)], and higher QOL scores (emotional function: 2.908; 95% CI: 0.600-5.216, P = 0.014; global health: 6.775; 95% CI: 3.967-9.583).

CONCLUSIONS: ePRO-based collaborative management effectively lessened postoperative burden and improved QOL beyond 6 months.

PMID:40775446 | DOI:10.1038/s41416-025-03110-5

Sci Rep. 2025 Aug 7;15(1):28949. doi: 10.1038/s41598-025-14320-5.

ABSTRACT

With the growing demand for English language teaching, the efficient and accurate evaluation of students’ writing ability has become a key focus in English education. This study introduces a Hybrid Feature-based Cross-Prompt Automated Essay Scoring (HFC-AES) model that leverages deep learning for intelligent text analysis. Building on traditional deep neural networks (DNNs), the model incorporates text structure features and attention mechanisms, while adversarial training is employed to optimize feature extraction and enhance cross-prompt adaptability. In the topic-independent stage, statistical methods and DNNs extract shared features for preliminary scoring. In the topic-specific stage, topic information is integrated into a hierarchical neural network to improve semantic understanding and topic alignment. Compared with existing Transformer-based scoring models, HFC-AES demonstrates superior robustness and semantic modeling capabilities. Experimental results show that HFC-AES achieves strong cross-prompt scoring performance, with an average Quadratic Weighted Kappa (QWK) of 0.856, outperforming mainstream models. Ablation studies further highlight the critical role of text structure features and attention mechanisms, particularly in improving argumentative writing assessment. Overall, HFC-AES offers effective technical support for automated essay grading, contributing to more reliable and efficient evaluation in English language teaching.

PMID:40775439 | DOI:10.1038/s41598-025-14320-5

Sci Rep. 2025 Aug 7;15(1):28990. doi: 10.1038/s41598-025-14276-6.

ABSTRACT

Straightlining (uniform responses across items), poses a risk in surveys. Among adolescents, previous studies have investigated the prevalence and impact of straightlining in shorter questionnaires within larger surveys. A typical finding is that straightlining is more common among younger respondents, and particularly among boys. A better understanding of straightlining is important for improving data quality. The present study aims to estimate the prevalence of straightlining among adolescents filling out a survey covering different aspects of social media use across 64 items. Additionally, it seeks to assess the impact of straightlining on internal consistency and the associations between six domains of social media use and symptoms of anxiety and depression. Data from the «LifeOnSoMe»-study (N = 3,285), collected from adolescents (aged 16+) in Bergen, Norway. Descriptive and inferential statistics. In total, 5.4% of participants were straightliners, (8.6% of the boys vs. 2.9% of the girls (p < 0.001)). There were no differences in age between straightliners and the remainder of the sample. Overall, the prevalence and impact of straightlining was limited in the present sample. However, there were large discrepancies in terms of both internal consistency, correlations between domains of social media use, and associations with symptoms of anxiety and depression between straightliners and the remainder of the sample. Straightlining behavior had minimal effects on this sample’s analytical epidemiological conclusions. While boys were more prone to straightlining than girls, overall prevalence was low. However, significant discrepancies between straightliners and other respondents suggest potential risks in samples with higher straightlining prevalence.

PMID:40775438 | DOI:10.1038/s41598-025-14276-6

Sci Rep. 2025 Aug 7;15(1):28969. doi: 10.1038/s41598-025-13240-8.

ABSTRACT

This study assesses the practical implications of ChatGPT’s ability to identify hotspots by comparing its performance to Geographical Information System (GIS) software in detecting CO₂ sources and sinks observed by the Orbiting Carbon Observatory-3 (OCO-3) satellite. ChatGPT exhibited performance comparable to ArcGIS in both z-score statistics and spatial distribution patterns of XCO₂ hot and cold spots. The results generated by ChatGPT showed a strong correlation with ArcGIS-generated hotspots, demonstrating a z-score correlation coefficient of R²=0.82 and a cosine similarity score of 0.90. As multimodal artificial intelligence becomes more prevalent in earth monitoring, ChatGPT is expected to be a valuable tool for identifying CO₂ emission patterns, particularly for users who lack specialized GIS expertise. These findings establish a significant benchmark for ChatGPT’s potential in this field, offering a novel approach to identifying area-wide spatial patterns of CO₂ emissions compared to conventional GIS software.

PMID:40775428 | DOI:10.1038/s41598-025-13240-8

Sci Rep. 2025 Aug 7;15(1):28925. doi: 10.1038/s41598-025-13380-x.

ABSTRACT

This study used data from a large dam site to model changes in groundwater quality variables. Several indicators were investigated to check the quality of water sources for the site for different purposes. The factor analysis results displayed that four factors control 87.58% of water quality changes. The primary factor responsible for approximately half of the impact on water quality, accounting for 55.12% of the total variance, includes the EC, Ca²⁺, SAR, SO₄^–, Na⁺, CO₃^–, %Na, Cl^–, and TDS parameters. These parameters are directly related to water quality and are influenced by the natural characteristics of the region. Considering that the main control factor for water quality is the first factor mentioned, these factors were used in multivariate analysis and intelligent modeling. Therefore, Na⁺, Cl⁺, Na%, CO₃^–, and SO₄^2- were used as input variables (independent variables), and EC, TDS, and SAR were used as output variables (dependent variables). Support vector machine (SVM) with various kernel functions, multilayer perceptron artificial neural network (MLP-ANN) with various training algorithms, random forest algorithm (RFA), Gaussian process regression (GPR), and statistical analysis methods were used for modeling. Among the kernel functions used in SVM, the radial basis function (RBF) kernel provided the most accurate results. On the other hand, among the learning algorithms used in neural networks, the Levenberg-Marquardt algorithm demonstrated the highest level of accuracy. Modeling results based on error value, Wilmot agreement index, A₂₀ index, determination coefficient, and violin diagrams showed that the SVM (R² > 0.99, RMSE < 0.04, A₂₀ = 1.00, WAI = 1.00) achieved better than the other models. The results of Kruskal-Wallis’s test disclosed that there is no substantial difference between the water quality parameters obtained from the models and the measured values.

PMID:40775421 | DOI:10.1038/s41598-025-13380-x

Neural Netw. 2025 Aug 5;192:107925. doi: 10.1016/j.neunet.2025.107925. Online ahead of print.

ABSTRACT

The development of efficient segmentation strategies for medical images has evolved from its initial dependence on Convolutional Neural Networks (CNNs) to the current investigation of hybrid models that combine CNNs with Vision Transformers (ViTs). There is an increasing focus on developing architectures that are both high-performing and computationally efficient, capable of being deployed on remote systems with limited resources. Although transformers can capture global dependencies in the input space, they face challenges from the corresponding high computational and storage expenses involved. The objective of this research is to propose that Vision Extended Long Short-Term Memory (Vision-xLSTM) forms an appropriate backbone for medical image segmentation, offering excellent performance with reduced computational costs. This study investigates the integration of CNNs with Vision-xLSTM by introducing the novel U-VixLSTM. The Vision-xLSTM blocks capture the temporal and global relationships within the patches extracted from the CNN feature maps. The convolutional feature reconstruction path upsamples the output volume from the Vision-xLSTM blocks to produce the segmentation output. The U-VixLSTM exhibits superior performance compared to the state-of-the-art networks in the publicly available Synapse, ISIC and ACDC datasets. The findings suggest that U-VixLSTM is a promising alternative to ViTs for medical image segmentation, delivering effective performance without substantial computational burden. This makes it feasible for deployment in healthcare environments with limited resources for faster diagnosis. Code provided: https://github.com/duttapallabi2907/U-VixLSTM.

PMID:40773779 | DOI:10.1016/j.neunet.2025.107925

JMIR Public Health Surveill. 2025 Aug 7;11:e72064. doi: 10.2196/72064.

ABSTRACT

BACKGROUND: Violence against women (VAW) is a major public health and human rights concern with profound mental health consequences. However, the association between specific VAW forms and mental health, particularly among left-behind women in rural China, remains unclear.

OBJECTIVE: This study aimed to identify the associations of VAW with depression, anxiety, and comorbid symptoms and to explore the potential roles of resilience and social support.

METHODS: The cross-sectional survey was conducted in Y City, Henan Province, China, in July 2023. A multistage stratified random sampling method was used to recruit left-behind women, resulting in a final sample of 1503 participants. Data on participants and their VAW were collected through a face-to-face questionnaire survey. The forms of VAW assessed were nonpartner violence (NPV) and intimate partner violence (IPV; including remote IPV). Depressive symptoms were evaluated using the 10-item Center for Epidemiological Studies Depression Scale, while anxiety symptoms were assessed with the Generalized Anxiety Disorder-7. The comorbid symptoms of depression and anxiety (CDA) were ascertained as the simultaneous presence of depressive and anxiety symptoms. A multivariable logistic regression model was used to estimate the odds ratio and 95% CIs. A 4-way decomposition analysis was conducted to test the mediation roles and interactions of resilience and social support between VAW and mental health outcomes. Population attributable fractions and pathway-specific population attributable fractions were calculated to estimate the burden of mental health outcomes attributable to VAW.

RESULTS: Lifetime VAW (adjusted odds ratio [aOR] 1.84, 95% CI 1.32-2.54) was associated with an increased risk of CDA. Women who were exposed to lifetime IPV (aOR 1.84, 95% CI 1.32-2.56), remote IPV (aOR 2.79, 95% CI 1.60-4.74), and NPV (aOR 2.63, 95% CI 1.58-4.26) had an increased likelihood of reporting CDA. Similar associations could also be found for depressive symptoms and anxiety symptoms. In the 4-way decomposition analysis for VAW and CDA, mediation effects of low resilience and social support were statistically significant (P<.05), whereas none of the interactions reached significance (P>.05). The pure mediation proportion was 28.2% for the low resilience and 18.6% for the social support between VAW and CDA. A total of 20.8% of CDA cases, 15.1% of depressive symptoms cases, and 22.7% of anxiety symptoms cases were attributable to VAW. Among these, low resilience accounted for 7.2% and low social support accounted for 4.7% of CDA cases as mediators.

CONCLUSIONS: Lifetime VAW, including IPV (and remote IPV) and NPV, shows significant associations with CDA and depressive and anxiety symptoms among rural left-behind women in China. The associations are partly mediated by low resilience and social support. Targeted strategies, including efforts to reduce violence against rural left-behind women, enhance their resilience and strengthen their social support networks, are urgently needed.

PMID:40773765 | DOI:10.2196/72064

Obstet Gynecol. 2025 Aug 7. doi: 10.1097/AOG.0000000000006028. Online ahead of print.

ABSTRACT

Healthy People 2030 aims to decrease low-risk cesarean delivery rates to 23.6% in the United States. In 2023, the national rate was 26.6%, though rates vary widely by state and hospital. This suggests a need for localized geographic estimates to identify places with higher burden. We modeled 2023 low-risk cesarean delivery rates by county of birth using birth certificate data and hierarchical Bayesian models that spatially smooth unstable estimates. We found considerable variation in rates, with county rates ranging from 5.8% to 53.4%. Counties in the West had lower rates than those in the Midwest, South, and Northeast. County rates increased with urbanicity. Only 47.7% (985) of counties had rates meeting the Healthy People 2030 target.

PMID:40773757 | DOI:10.1097/AOG.0000000000006028