Automatic Graph Topology-Aware Transformer

Existing efforts are dedicated to designing many topologies and graph-aware strategies for the graph Transformer, which greatly improve the model's representation capabilities. However, manually determining the suitable Transformer architecture for a specific graph dataset or task requires extensive expert knowledge and laborious trials. This article proposes an evolutionary graph Transformer architecture search (EGTAS) framework to automate the construction of strong graph Transformers. We build a comprehensive graph Transformer search space with the micro-level and macro-level designs. EGTAS evolves graph Transformer topologies at the macro level and graph-aware strategies at the micro level. Furthermore, a surrogate model based on generic architectural coding is proposed to directly predict the performance of graph Transformers, substantially reducing the evaluation cost of evolutionary search. We demonstrate the efficacy of EGTAS across a range of graph-level and node-level tasks, encompassing both small-scale and large-scale graph datasets. Experimental results and ablation studies show that EGTAS can construct high-performance architectures that rival state-of-the-art manual and automated baselines.

Mortality prediction after major surgery in a mixed population through machine learning: a multi-objective symbolic regression approach

INTRODUCTION

Understanding 1-year mortality following major surgery offers valuable insights into patient outcomes and the quality of peri-operative care. Few models exist that predict 1-year mortality accurately. This study aimed to develop a predictive model for 1-year mortality in patients undergoing complex non-cardiac surgery using a novel machine-learning technique called multi-objective symbolic regression.

METHODS

A single-institution database of patients undergoing major elective surgery with previous cardiopulmonary exercise testing was divided into three datasets: pre-operative clinical data; cardiorespiratory and physiological data; and combined. A multi-objective symbolic regression model was developed and compared against existing models. Model performance was evaluated using the F1 score. Shapley additive explanations analysis was used to identify the major contributors to model performance.

RESULTS

From 2145 patients in the database, 1190 were included, with 952 in the training dataset and 238 in the test dataset. Median (IQR [range]) age was 71 (61-79 [45-89]) years and 825 (69%) were male. The multi-objective symbolic regression model demonstrated robust consistency with an F1 score of 0.712. Shapley additive explanations analysis indicated that ventilatory equivalents for carbon dioxide, oxygen at peak exercise and BMI influenced model performance most significantly, surpassing surgery type and named comorbidities.

DISCUSSION

This study confirms the feasibility of developing a multi-objective symbolic regression-based model for predicting 1-year postoperative mortality in a mixed non-cardiac surgical population. The model's strong performance underscores the critical role of physiological data, particularly cardiorespiratory fitness, in surgical risk assessment and emphasises the importance of pre-operative optimisation to identify and manage high-risk patients. The multi-objective symbolic regression model demonstrated high sensitivity and a good F1 score, highlighting its potential as an effective tool for peri-operative risk prediction.

Principles for enhancing trust in artificial intelligence systems among medical imaging professionals in Ghana: A nationwide cross-sectional study

INTRODUCTION

To realise the full potential of artificial intelligence (AI) systems in medical imaging, it is crucial to address challenges, such as cyberterrorism to foster trust and acceptance. This study aimed to determine the principles that enhance trust in AI systems from the perspective of medical imaging professionals in Ghana.

METHODS

An anonymous, online, nationwide cross-sectional survey was conducted. The survey contained questions related to socio-demographic characteristics and AI trustworthy principles, including "human agency and oversight", "technical robustness and safety", "data privacy, security and governance" and "transparency, fairness and accountability".

RESULTS

A total of 370 respondents completed the survey. Among the respondents, 66.5 % (n = 246) were diagnostic radiographers. Considerable number of respondents (n = 121, 32.7 %) reported having little or no understanding of how medical imaging AI systems work. Overall, 54.9 % (n = 203) of the respondents agreed or strongly agreed that each of the four principles was important to enhance trust in medical imaging AI systems, with a composite mean score of 3.88 ± 0.45. Transparency, fairness and accountability had the highest rating (4.27 ± 0.58), whereas the mean score for human agency and oversight was 3.89 ± 0.53. Technical robustness and safety as well as data privacy, security and governance obtained mean scores of 3.79 ± 0.61 and 3.58 ± 0.65, respectively.

CONCLUSION

Medical imaging professionals in Ghana agreed that human agency, technical robustness, data privacy and transparency are important principles to enhance trust in AI systems; however, future plans including medical imaging AI educational interventions are required to improve AI literacy among medical imaging professionals in Ghana.

IMPLICATIONS FOR PRACTICE

The evidence presented should encourage organisations to design and deploy trustworthy medical imaging AI systems.

Elucidating ozone formation mechanisms in the central Yangtze River Delta region, China: Urban and rural differences

Surface ozone (O3) pollution has become a pressing air quality issue in eastern China in recent years. However, studies comparing O3 formation in urban and rural areas remain limited. This study presents a field campaign focusing on volatile organic compounds (VOCs) conducted at two sites in the central Yangtze River Delta (YRD) region during the warm season (June to August) of 2023. VOC pollution sources identified through positive matrix factorization (PMF) were integrated into a machine learning framework, along with nitrogen dioxide (NO2) and meteorological factors, to quantify their impacts on O3 formation. The results show that urban areas have higher VOC concentrations, primarily driven by elevated levels of aromatics and oxygenated volatile organic compounds (OVOCs), compared to rural areas. PMF analysis identified six major VOC sources: industrial emissions, paint and solvent usage, biogenic emissions, combustion-related emissions, mobile sources, and liquefied petroleum gas usage. Mobile sources and industrial emissions are more significant in urban areas, while combustion-related emissions are more significant in rural areas. The machine learning model effectively captured the relationships between meteorological parameters, precursors, and O3 levels. Analysis revealed that meteorological factors are the primary drivers of O3 formation in rural areas, whereas both meteorological factors and precursors contribute equally in urban areas. Relative humidity and combustion source emerged as the most influential factors at both sites, though the significance of other factors varied due to environmental differences. These findings enhance our understanding of O3 pollution differences between urban and rural regions. The combined effects of meteorological factors, NO2, and VOCs should be taken into account when formulating O3 pollution control policies.

Factors that affect clinical youth engagement in digital mental health research: a qualitative sub-study nested within a prospective cohort study

BACKGROUND

There has been extensive debate about the role of social media and smartphone use in youth mental health and self-harm. Research to date lacks sufficient detail to determine the mechanisms underpinning any associations. The Social Media, Smartphone use and Self-harm in Young People (3S-YP) study is a prospective cohort study that was co-produced with young people to investigate temporal patterns of social media and smartphone use prior to an episode of self-harm in a clinical youth sample. Young people were actively involved in all key stages of the research process to ensure the research would be relevant and acceptable to the intended population. This included defining the research question and designing the methods. This qualitative sub-study nested within the main 3S-YP study aimed to evaluate young people's experiences of engaging in this innovative digital mental health study. This will help inform understanding regarding the added value of co-production and future research in this field.

METHODS

Semi-structured interviews were conducted with a purposive sample of participants from the 3S-YP study. Interview data was analysed using codebook thematic analysis.

RESULTS

Sixteen young people (mean 19.8 years old, SD 2.9; n = 10 female, 63%) participated in the interviews. Participants were generally comfortable answering questions about sensitive topics using remote digital tools, appreciating the greater privacy, convenience and opportunity for self-reflection they provide, whilst noting periods of poor mental health may affect study engagement. The remote research methods (including the participation information and tools for recruitment and data collection) were considered user-friendly and were complemented by the active role of the research team who facilitated young people's engagement with the study. Despite the relevance and support for research on the impact of digital technology use on youth mental health, concerns about data sharing and a complex process for accessing data from social media platforms complicated study engagement. The role of parental involvement was also described.

CONCLUSIONS

User-friendly remote research methods, coupled with proactive, responsive researchers and parental support are beneficial for conducting research with clinical youth populations. Whilst young people endorse research in this field, concerns about data sharing and barriers to data access need addressing if researchers are to effectively employ innovative solutions to investigating the impact of smartphones and social media use on youth mental health and self-harm. The findings from this study demonstrate the value of actively involving those with lived experience throughout the research process and provide useful insight for researchers intending to conduct similar research.

TRIAL REGISTRATION

This study is registered on ClinicalTrials.gov (ID no. NCT04601220).

Investigating Protective and Risk Factors and Predictive Insights for Aboriginal Perinatal Mental Health: Explainable Artificial Intelligence Approach

BACKGROUND

Perinatal depression and anxiety significantly impact maternal and infant health, potentially leading to severe outcomes like preterm birth and suicide. Aboriginal women, despite their resilience, face elevated risks due to the long-term effects of colonization and cultural disruption. The Baby Coming You Ready (BCYR) model of care, centered on a digitized, holistic, strengths-based assessment, was co-designed to address these challenges. The successful BCYR pilot demonstrated its ability to replace traditional risk-based screens. However, some health professionals still overrely on psychological risk scores, often overlooking the contextual circumstances of Aboriginal mothers, their cultural strengths, and mitigating protective factors. This highlights the need for new tools to improve clinical decision-making.

OBJECTIVE

We explored different explainable artificial intelligence (XAI)-powered machine learning techniques for developing culturally informed, strengths-based predictive modeling of perinatal psychological distress among Aboriginal mothers. The model identifies and evaluates influential protective and risk factors while offering transparent explanations for AI-driven decisions.

METHODS

We used deidentified data from 293 Aboriginal mothers who participated in the BCYR program between September 2021 and June 2023 at 6 health care services in Perth and regional Western Australia. The original dataset includes variables spanning cultural strengths, protective factors, life events, worries, relationships, childhood experiences, family and domestic violence, and substance use. After applying feature selection and expert input, 20 variables were chosen as predictors. The Kessler-5 scale was used as an indicator of perinatal psychological distress. Several machine learning models, including random forest (RF), CatBoost (CB), light gradient-boosting machine (LightGBM), extreme gradient boosting (XGBoost), k-nearest neighbor (KNN), support vector machine (SVM), and explainable boosting machine (EBM), were developed and compared for predictive performance. To make the black-box model interpretable, post hoc explanation techniques including Shapley additive explanations and local interpretable model-agnostic explanations were applied.

RESULTS

The EBM outperformed other models (accuracy=0.849, 95% CI 0.8170-0.8814; F1-score=0.771, 95% CI 0.7169-0.8245; area under the curve=0.821, 95% CI 0.7829-0.8593) followed by RF (accuracy=0.829, 95% CI 0.7960-0.8617; F1-score=0.736, 95% CI 0.6859-0.7851; area under the curve=0.795, 95% CI 0.7581-0.8318). Explanations from EBM, Shapley additive explanations, and local interpretable model-agnostic explanations identified consistent patterns of key influential factors, including questions related to "Feeling Lonely," "Blaming Herself," "Makes Family Proud," "Life Not Worth Living," and "Managing Day-to-Day." At the individual level, where responses are highly personal, these XAI techniques provided case-specific insights through visual representations, distinguishing between protective and risk factors and illustrating their impact on predictions.

CONCLUSIONS

This study shows the potential of XAI-driven models to predict psychological distress in Aboriginal mothers and provide clear, human-interpretable explanations of how important factors interact and influence outcomes. These models may help health professionals make more informed, non-biased decisions in Aboriginal perinatal mental health screenings.

Clinical assessment of the criticality index - dynamic, a machine learning prediction model of future care needs in pediatric inpatients

OBJECTIVE

To assess patient characteristics and care factors that are associated with correct and incorrect predictions of future care locations (ICU vs. non-ICU) by the Criticality Index-Dynamic (CI-D), with the goal of enhancing the CI-D.

DESIGN

Retrospective structured chart review.

PARTICIPANTS

All pediatric inpatients admitted from January` 1st 2018 - February 29th 2020 through the emergency department.

MAIN OUTCOME(S) AND MEASURE(S)

Patient characteristics and care factors associated with correct (true positives, true negatives) and incorrect predictions (false positives, false negatives) of future care locations (ICU vs. non-ICU) by the CI-D were assessed.

RESULTS

Of the 3,018, patients, 139 transitioned from non-ICU locations to ICU care; 482 were transferred from the ICU to non-ICU care locations, and 2,400 remained in non-ICU care locations. For the ICU Prediction group, the false negative patients were older, more frequently male, and had longer hospital and ICU lengths of stay compared to the true positive patients. The significant differences in the ICU Prediction group for false negative compared to the true positive patients included a less frequent: primary diagnosis of respiratory failure, use of high flow nasal canula, hourly cardio-respiratory vital signs prior to transfer to the ICU, and neurologic vital signs after transfer from the ICU. For the ICU Discharge prediction group, false positive patients were more frequently: younger, had a primary diagnosis of respiratory failure, more frequently received respiratory support after discharge from the ICU, and received less frequent neurological vital signs prior to transfer from the ICU. For the Non-transfer prediction category, demographics and clinical variables did not differ between the true negative and false positive prediction groups.

CONCLUSION AND RELEVANCE

We conducted the first comprehensive analysis via structured chart reviews of patient characteristics and care factors that are associated with correct and incorrect predictions of future care locations by the machine learning algorithm, the CI-D, gaining insights into potential new predictor variables for inclusion in the model to improve future model iterations.

TF-LIME : Interpretation Method for Time-Series Models Based on Time-Frequency Features

With the widespread application of machine learning techniques in time series analysis, the interpretability of models trained on time series data has attracted increasing attention. Most existing explanation methods are based on time-domain features, making it difficult to reveal how complex models focus on time-frequency information. To address this, this paper proposes a time-frequency domain-based time series interpretation method aimed at enhancing the interpretability of models at the time-frequency domain. This method extends the traditional LIME algorithm by combining the ideas of short-time Fourier transform (STFT), inverse STFT, and local interpretable model-agnostic explanations (LIME), and introduces a self-designed TFHS (time-frequency homogeneous segmentation) algorithm. The TFHS algorithm achieves precise homogeneous segmentation of the time-frequency matrix through peak detection and clustering analysis, incorporating the distribution characteristics of signals in both frequency and time dimensions. The experiment verified the effectiveness of the TFHS algorithm on Synthetic Dataset 1 and the effectiveness of the TF-LIME algorithm on Synthetic Dataset 2, and then further evaluated the interpretability performance on the MIT-BIH dataset. The results demonstrate that the proposed method significantly improves the interpretability of time-series models in the time-frequency domain, exhibiting strong generalization capabilities and promising application prospects.

Mobile Phone Network Data in the COVID-19 era: A systematic review of applications, socioeconomic factors affecting compliance to non-pharmaceutical interventions, privacy implications, and post-pandemic economic recovery strategies

BACKGROUND

The use of traditional mobility datasets, such as travel surveys and census data, has significantly impacted various disciplines, including transportation, urban sensing, criminology, and healthcare. However, because these datasets represent only discrete instances of measurement, they miss continuous temporal shifts in human activities, failing to record the majority of human mobility patterns in real-time. Bolstered by the rapid expansion of telecommunication networks and the ubiquitous use of smartphones, mobile phone network data (MPND) played a pivotal role in fighting and controlling the spread of COVID-19.

METHODS

We conduct an extensive review of the state-of-the-art and recent advancements in the application of MPND for analyzing the early and post-stages of the COVID-19 pandemic, following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Additionally, we evaluate and assess the included studies using the Mixed Methods Appraisal Tool (MMAT) and the Critical Appraisal Skills Programme (CASP). Furthermore, we apply bibliometric analysis to visualize publication structures, co-authorship networks, and keyword co-occurrence networks.

RESULTS

After the full-text screening process against the inclusion and exclusion criteria, our systematic literature review identified 55 studies that utilized MPND in the context of the COVID-19 pandemic: 46 (83.6%) were quantitative, and 9 (16.4%) were qualitative. These quantitative studies can be classified into five main groups: monitoring and tracking of human mobility patterns (n = 11), investigating the correlation between mobility patterns and the spread of COVID-19 (n = 7), analyzing the recovery of economic activities and travel patterns (n = 5), assessing factors associated with NPI compliance (n = 5), and investigating the impact of COVID-19 lockdowns and non-pharmaceutical interventions (NPI) measures on human behaviors, urban dynamics, and economic activity (n = 18). In addition, our findings indicate that NPI measures had a significant impact on reducing human movement and dynamics. However, demographics, political party affiliation, socioeconomic inequality, and racial inequality had a significant impact on population adherence to NPI measures, which could increase disease spread and delay social and economic recovery.

CONCLUSION

The usage of MPND for monitoring and tracking human activities and mobility patterns during the COVID-19 pandemic raises privacy implications and ethical concerns. Thus, striking a balance between meeting the ethical requirements and maintaining privacy risks should be further discovered and investigated in the future.

A novel near real-time approach to forecast high resolution NO2 concentrations in southeastern China by incorporating multi-source satellite data

Nitrogen dioxide, designated as NO2, is a critical yet harmful trace gas in Earth's atmospheric composition. NO2 poses significant threats to human health, ecosystems, and agricultural productivity. Accurate NO2 forecasts at high spatial resolution enable authorities to safeguard public health through targeted mitigation efforts. Conventional NO2 forecasting approaches, such as time series analysis and chemical transport models (CTMs), often suffer from significant uncertainty or lack fine spatial details. This study presents a novel NO2 forecast model that combines Random Forest techniques with multi-source satellite data and NASA's Goddard Earth Observing System "Composing Forecasting" (GEOS-CF) product to provide spatially continuous, five-day forecasts of NO2 concentrations at 1 km resolution across southeastern China. The superior capabilities of our forecast framework were confirmed through multiple validation methods, consistently surpassing the performance of the original GEOS-CF model. Notably, the new framework achieved substantial error reductions and resolution enhancements in GEOS-CF forecasts, outperforming the initial product across all validation metrics. The developed model facilitates the generation of NO2 forecasts characterized by near-real-time delivery, great precision, and high spatial resolution.

Health-Promoting Effects and Everyday Experiences With a Mental Health App Using Ecological Momentary Assessments and AI-Based Ecological Momentary Interventions Among Young People: Qualitative Interview and Focus Group Study

BACKGROUND

Considering the high prevalence of mental health conditions among young people and the technological advancements of artificial intelligence (AI)-based approaches in health services, mobile health (mHealth) apps for mental health are a promising way for low-threshold and large-scale mental health promotion, prevention, and intervention strategies, especially for young people. However, insufficient evidence on health-promoting effects and deficient user-centric designs emphasize the necessity for participatory methods in the interventions' development processes.

OBJECTIVE

This study aimed to explore young people's everyday experiences using an AI-based mHealth app for mental health promotion based on ecological momentary assessments and ecological momentary interventions. Our analysis of qualitative data focused on exploring young people's use patterns in daily life and mental health-promoting effects.

METHODS

We conducted problem-centered interviews and focus groups with a subsample of 27 young people aged 14 to 25 years, who were among the participants of 2 microrandomized trials testing and evaluating an AI-based mHealth app (AI4U training). Our study used a participatory approach, with "co- and peer researchers" from the dialogue population actively engaged in research processes and data analysis. Structural content analysis guided the qualitative analysis.

RESULTS

Participants reported enhanced emotional self-awareness and regulation in daily life through the ecological momentary assessments and ecological momentary interventions. Young people appreciated the AI4U training for managing emotions and stress. They had no trust issues regarding disclosing their mental health via the AI4U training in daily life. Some faced challenges integrating it into their daily routines and highlighted the value of autonomy in use decision-making processes.

CONCLUSIONS

Our findings reveal that young people benefited from enhanced emotional awareness and management through the use of the AI4U training, appreciating its anonymity for facilitating emotional disclosure. The results suggest that enhanced self-directed use may improve daily life integration, although participants noted that they sometimes avoided using the AI4U training during distress despite recognizing its potential benefits. These findings indicate the importance of balancing directed use and autonomy in digital interventions to harmonize compliance with effectiveness in daily life. We highlight the importance of participatory research for tailored digital mental health solutions.

How does research output and impact in medical informatics vary among EU member states? - A bibliometric analysis

OBJECTIVES

This study evaluates how research output and impact in medical informatics vary among EU member states before and during the COVID-19 pandemic by analyzing publication volume, impact metrics, collaboration patterns, and open-access trends. It seeks to identify regional disparities, highlight key research themes, and provide insights for researchers, the public, and policymakers to promote equitable access, collaboration, and investment in medical informatics across the EU.

METHODS

A bibliometric analysis was performed using Clarivate Web of Science and InCites databases, encompassing 6,620 articles from 47 medical informatics journals published between 2018 and 2022. Metrics such as cumulative impact factors, article counts, and collaboration trends were analyzed.

RESULTS

Our analysis identified substantial regional disparities in research output and impact. Western European countries, including Germany, the Netherlands, and Spain, consistently led in article volume and cumulative impact factors, while Eastern European countries showed lower engagement. Collaboration metrics revealed that 66 % of publications involved international partnerships, showcasing strong cross-border cooperation within the EU.

CONCLUSIONS

This study highlights the uneven distribution of research productivity in medical informatics across the EU. The findings underline the importance of international partnerships and equitable access to research in advancing medical informatics and addressing evolving healthcare challenges.

Trackformers: in search of transformer-based particle tracking for the high-luminosity LHC era

High-Energy Physics experiments are facing a multi-fold data increase with every new iteration. This is certainly the case for the upcoming High-Luminosity LHC upgrade. Such increased data processing requirements forces revisions to almost every step of the data processing pipeline. One such step in need of an overhaul is the task of particle track reconstruction, a.k.a., tracking. A Machine Learning-assisted solution is expected to provide significant improvements, since the most time-consuming step in tracking is the assignment of hits to particles or track candidates. This is the topic of this paper. We take inspiration from large language models. As such, we consider two approaches: the prediction of the next word in a sentence (next hit point in a track), as well as the one-shot prediction of all hits within an event. In an extensive design effort, we have experimented with three models based on the Transformer architecture and one model based on the U-Net architecture, performing track association predictions for collision event hit points. In our evaluation, we consider a spectrum of simple to complex representations of the problem, eliminating designs with lower metrics early on. We report extensive results, covering both prediction accuracy (score) and computational performance. We have made use of the REDVID simulation framework, as well as reductions applied to the TrackML data set, to compose five data sets from simple to complex, for our experiments. The results highlight distinct advantages among different designs in terms of prediction accuracy and computational performance, demonstrating the efficiency of our methodology. Most importantly, the results show the viability of a one-shot encoder-classifier based Transformer solution as a practical approach for the task of tracking.

Machine Learning-Based Non-Invasive Prediction of Metabolic Dysfunction-Associated Steatohepatitis in Obese Patients: A Retrospective Study

Objectives: We aimed to develop and validate machine learning (ML) models that integrate clinical and laboratory data for the non-invasive prediction of metabolic dysfunction-associated steatohepatitis (MASH) in an obese population. Methods: In this retrospective study, clinical and laboratory data were collected from obese patients undergoing bariatric surgery. The cohort was divided using stratified random sampling, and optimal features were selected with SHapley Additive exPlanations (SHAP). Various ML models, including K-nearest neighbors, linear support vector machine, radial basis function support vector machine, Gaussian process, random forest, multilayer perceptron, adaptive boosting, and naïve Bayes, were developed through cross-validation and hyperparameter tuning. Diagnostic performance was assessed via the area under the curve (AUC) in both training and validation sets. Results: A total of 558 patients were analyzed, with 390 in the training set and 168 in the validation set. In the training cohort, the median age was 35 years, the median body mass index (BMI) was 39.8 kg/m2, 39.0% were male, 37.9% had diabetes mellitus, and 62.8% were diagnosed with MASH. The validation cohort had a median age of 34.1 years, a median BMI of 42.5 kg/m2, 41.7% male, 32.7% with diabetes, and 39.9% with MASH. Among the models, the random forest achieved the highest performance among the models with AUC values of 0.94 in the training set and 0.88 in the validation set. The Gaussian process model attained an AUC of 0.97 in the training cohort but 0.79 in the validation cohort, while the other models achieved AUC values ranging from 0.63 to 0.88 in the training cohort and 0.62 to 0.75 in the validation set. Conclusions: ML models, particularly the random forest, effectively predict MASH using readily available data, offering a promising non-invasive alternative to conventional serological scoring. Prospective studies and external validations are needed to further establish clinical utility.

Artificial intelligence in inflammatory bowel disease

ABSTRACT

Inflammatory bowel disease (IBD) is a complex condition influenced by various intestinal factors. Advances in next-generation sequencing, high-throughput omics, and molecular network technologies have significantly accelerated research in this field. The emergence of artificial intelligence (AI) has further enhanced the efficient utilization and interpretation of datasets, enabling the discovery of clinically actionable insights. AI is now extensively applied in gastroenterology, where it aids in endoscopic analyses, including the diagnosis of colorectal cancer, precancerous polyps, gastrointestinal inflammatory lesions, and bleeding. Additionally, AI supports clinicians in patient stratification, predicting disease progression and treatment responses, and adjusting treatment plans in a timely manner. This approach not only reduces healthcare costs but also improves patient health and safety. This review outlines the principles of AI, the current research landscape, and future directions for its applications in IBD, with the goal of advancing targeted treatment strategies.

Comparison of dynamic mode decomposition with other data-driven models for lung cancer incidence rate prediction

Introduction

Public health data analysis is critical to understanding disease trends. Existing analysis methods struggle with the complexity of public health data, which includes both location and time factors. Machine learning offers powerful tools but can be computationally expensive and require specialized knowledge. Dynamic mode decomposition (DMD) is an alternative that offers efficient analysis with fewer resources. This study explores applying DMD in public health using lung cancer data and compares it with other machine learning models.

Methods

We analyzed lung cancer incidence data (2000-2021) from 1,013 US counties. Machine learning models (random forest, gradient boosting machine, support vector machine) were trained and optimized on the training data. We also employed time series, a linear regression model, and DMD for comparison. All models were evaluated based on their ability to predict 2021 lung cancer incidence rates.

Results

The time series model achieved the lowest root mean squared error, followed by random forest. Meanwhile, DMD had an RMSE similar to that of Random Forest. Nearly all counties in Kentucky had higher lung cancer incidence rates, while states like California, New Mexico, Utah, and Idaho showed lower trends.

Conclusion

In summary, DMD offers a promising alternative for public health professionals to capture underlying trends and potentially have lower computational demands compared to other machine learning models.

Artificial intelligence in the diagnosis and management of gynecologic cancer

Gynecologic cancers affect over 1.2 million women globally each year. Early diagnosis and effective treatment are essential for improving patient outcomes, yet traditional diagnostic methods often encounter limitations, particularly in low-resource settings. Artificial intelligence (AI) has emerged as a transformative tool that enhances accuracy and efficiency across various aspects of gynecologic oncology, including screening, diagnosis, and treatment. This review examines the current applications of AI in gynecologic cancer care, focusing on areas such as early detection, imaging, personalized treatment planning, and patient monitoring. Based on an analysis of 75 peer-reviewed articles published between 2017 and 2024, we highlight AI's contributions to cervical, ovarian, and endometrial cancer management. AI has notably improved early detection, achieving up to 95% accuracy in cervical cancer screening through AI-enhanced Pap smear analysis and colposcopy. For ovarian and endometrial cancers, AI-driven imaging and biomarker detection have enabled more personalized treatment approaches. In addition, AI tools have enhanced precision in robotic-assisted surgery and radiotherapy, and AI-based histopathology has reduced diagnostic variability. Despite these advancements, challenges such as data privacy, bias, and the need for human oversight must be addressed. The successful integration of AI into clinical practice will require careful consideration of ethical issues and a balanced approach that incorporates human expertise. Overall, AI presents significant potential to improve outcomes in gynecologic oncology, particularly in bridging healthcare gaps in resource-limited settings.

UKB-MDRMF: a multi-disease risk and multimorbidity framework based on UK biobank data

The rapid accumulation of biomedical cohort data presents opportunities to explore disease mechanisms, risk factors, and prognostic markers. However, current research often has a narrow focus, limiting the exploration of risk factors and inter-disease correlations. Additionally, fragmented processes and time constraints can hinder comprehensive analysis of the disease landscape. Our work addresses these challenges by integrating multimodal data from the UK Biobank, including basic, lifestyle, measurement, environment, genetic, and imaging data. We propose UKB-MDRMF, a comprehensive framework for predicting and assessing health risks across 1560 diseases. Unlike single disease models, UKB-MDRMF incorporates multimorbidity mechanisms, resulting in superior predictive accuracy, with all disease types showing improved performance in risk assessment. By jointly predicting and assessing multiple diseases, UKB-MDRMF uncovers shared and distinctive connections among risk factors and diseases, offering a broader perspective on health and multimorbidity mechanisms.

Explainable machine learning framework for biomarker discovery by combining biological age and frailty prediction

Biological age (BA) and frailty represent two distinct health measures that offer valuable insights into the aging process. Comparing and analyzing blood-based biomarkers from the machine learning (ML) predictors of BA and frailty helps deepen our understanding of aging. This study aimed to develop a novel framework to identify biomarkers of aging by combining BA and frailty ML predictors with eXplainable Artificial Intelligence (XAI) techniques. We utilized data from middle-aged and older Chinese adults (≥ 45 years) in the 2011/2012 wave (n = 9702) and the 2015/2016 wave (n = 9455, as test set validation) of the China Health and Retirement Longitudinal Study (CHARLS). Sixteen blood-based biomarkers were used to predict BA and frailty. Four tree-based ML algorithms were employed in the training and validation, and performance metrics were compared to select the best models. Then, SHapley Additive exPlanations (SHAP) analysis was conducted on the selected models. CatBoost performed the best in the BA predictor, and Gradient Boosting performed the best in the frailty predictor. Traditional ML feature importance identified cystatin C and glycated hemoglobin as the major contributors for their respective models. However, subsequent SHAP analysis demonstrated that only cystatin C was the primary contributor in both models. The proposed framework can easily incorporate additional biomarkers, providing a scalable and comprehensive toolset that offers a quantitative understanding of biomarkers of aging.

Predicting mortality and risk factors of sepsis related ARDS using machine learning models

Sepsis related acute respiratory distress syndrome (ARDS) is a common and serious disease in clinic. Accurate prediction of in-hospital mortality of patients is crucial to optimize treatment and improve prognosis under the new global definition of ARDS. Our study aimed to use machine learning models to develop models that can effectively predict the in-hospital mortality of patients with sepsis related ARDS, calculate the mortality, and to identify related risk factors under the new global definition of ARDS. Based on MIMIC database, our study included 3470 first-time admission records of patients with sepsis related ARDS. After excluding 4 patients under the age of 18, 75 patients with less than 24 h stay in ICU, and 5 cases with missing indicators > 30%, finally 3386 cases were retained. The variance inflation factor (VIF) analysis was used to test the collinearity of the explanatory variables. The data were divided into the training set and the test set according to the ratio of 7:3. Six models, extreme gradient boosting (XGBoost), light gradient boosting (LightGBM), random forest (RF), classification and regression tree (CART), naive bayes (NB) and logistic regression (LR), were designed for training and testing. In the training set, XGBoost (AUROC = 0.951, 95% CI 0.942-0.961), LR (AUROC = 0.835, 95% CI 0.817-0.854), RF (AUROC = 1.0, 95% CI 1.0-1.0), LightGBM (AUROC = 1.0, 95% CI 1.0-1.0), CART (AUROC = 0.831, 95% CI 0.811-0.852), NB (AUROC = 0.793, 95% CI 0.772-0.814). In the test set, XGBoost (AUROC = 0.833, 95% CI 0.804-0.861), LR (AUROC = 0.82695% CI 0.796-0.856), RF (AUROC = 0.846, 95% CI 0.818-0.874), LightGBM (AUROC = 0.827, 95% CI 0.798-0.856), CART (AUROC = 0.753, 95% CI 0.718-0.787), NB (AUROC = 0.799, 95% CI 0.768-0.831). The RF model has the best performance on the test set. Further analyze the feature importance ranking and partial dependence plots of random forest model. Acute physiology and chronic health evaluation III (APACHE III), bicarbonate, anion gap and non-invasive blood pressure systolic were identified as the four most important risk characteristics. In this study, a variety of machine learning models have been successfully constructed to predict the in-hospital mortality of patients with sepsis related ARDS, among which the RF model performs well. Key risk factors identified include APACHE III, bicarbonate, anion gap and non-invasive blood pressure systolic. The identification of these factors helps clinicians to assess patients' conditions more accurately and develop personalized treatment plans, thereby improving the survival rate and prognosis quality of patients under the new global definition of ARDS.

Explanation and elaboration of MedinAI: guidelines for reporting artificial intelligence studies in medicines, pharmacotherapy, and pharmaceutical services

The increasing adoption of artificial intelligence (AI) in medicines, pharmacotherapy, and pharmaceutical services necessitates clear guidance on reporting standards. While the MedinAI Statement (Bottacin in Int J Clin Pharm, https://doi.org/10.1007/s11096-025-01905-3, 2025) provides core guidelines for reporting AI studies in these fields, detailed explanations and practical examples are crucial for optimal implementation. This companion document was developed to offer comprehensive guidance and real-world examples for each guideline item. The document elaborates on all 14 items and 78 sub-items across four domains: core, ethical considerations in medication and pharmacotherapy, medicines as products, and services related to medicines and pharmacotherapy. Through clear, actionable guidance and diverse examples, this document enhances MedinAI's utility, enabling researchers and stakeholders to improve the quality and transparency of AI research reporting across various contexts, study designs, and development stages.

Analysis of a medication discrepancy management platform in reducing medication discrepancy and influencing factors among elderly patients with polypharmacy

OBJECTIVE

This study aimed to investigate the impact of a medication discrepancy management platform on reducing medication discrepancies among elderly patients with polypharmacy and to analyze influencing factors.

METHODS

A total of 110 elderly polypharmacy patients were divided into a control group and an observation group using a random number method, each with 55 participants. The control group received routine management, while the observation group utilized a medication discrepancy management platform. Medication knowledge and adherence before and after intervention were compared between the two groups. Reasons and types of medication discrepancies were statistically analyzed. Patients were divided into a non-discrepancy group and a discrepancy group, with multivariate logistic regression used to analyze factors influencing medication discrepancies among elderly patients with polypharmacy.

RESULTS

Utilizing a medication discrepancy management platform significantly improved medication knowledge and adherence among elderly patients (P < 0.05). A total of 34 patients (30.91%) experienced at least one medication discrepancy within one-week post-discharge, primarily involving decreased frequency, missed doses, reduction in medication types, and medication substitution. Multivariate logistic regression analysis showed that the use of the medication discrepancy management platform, caregiver involvement, and prescribed discharge medications (7-8 types or ≥ 9 types) were independent factors influencing medication discrepancies in elderly patients (P < 0.05).

CONCLUSION

Using a medication discrepancy management platform can effectively reduce medication discrepancies in elderly patients with polypharmacy and improve elderly patients' adherence to medication. Expanding the platform's use can enhance discharge guidance quality and ensure medication safety.

Impacts of Inorganic Arsenic Exposure on Genetic Stability of Human Mesenchymal Stromal Cells

Human mesenchymal stem/stromal cells (hMSCs) can differentiate into mesoderm-type cells, making them suitable candidates for tissue repair therapies. However, their relatively low frequency in adult tissue necessitates ex vivo expansion prior to regenerative medicine applications, and therefore, long-term hMSC genetic stability during expansion should be studied. hMSC applications in regenerative medicine ensure commercial availability of normal karyotype human primary cells for toxicity assessment and hMSCs could serve as alternatives to immortalized human cell models. In this work, we evaluated the potential of hMSCs in toxicity testing using inorganic arsenic (iAs) as a case study. hMSCs were exposed to iAs at different durations to track cellular aging and study long-term genetic stability. iAs exposures (48 h) resulted in micronuclei induction. hMSCs were also exposed to iAs for 6 days to determine if hMSCs would become more susceptible to chromosomal damage following exposure to the model genotoxicant, mitomycin C (MMC). The culture duration and iAs exposure did not alter MMC potency, indicating that the hMSC susceptibility to chromosomal damage remained unchanged. We also used gene expression analysis to investigate the molecular impacts of iAs on hMSCs over the course of short (3 days total) and long (30 days total) experiments. Both iAs exposures activated biomarkers associated with oxidative stress, but not biomarkers for direct DNA damage, providing support for an indirect mode of action for iAs genotoxicity. Overall, this study establishes the utility of hMSCs as a new model for toxicity screening and provides mechanistic information underlying iAs toxicity.

Discovery of Dynamic Models for AML Disease Progression from Longitudinal Multi-Modal Clinical Data Using Explainable Machine Learning

Acute Myeloid Leukemia (AML) is a complex and heterogeneous disease identified by severe clinical progression, fast cellular proliferation, and often high mortality rates. Incorporating diverse longitudinal information on patients' medical histories is essential for developing effective disease predictive models applicable to both research and clinical settings. Here, we present a robust methodology for discovering dynamic predictive models to elucidate AML disease progression dynamics from a novel longitudinal multimodal clinical dataset of patients diagnosed with AML. The clinical dataset was analyzed to reveal the main clinical, genetic, and treatment features modulating disease progression. To discover mathematical models-including interactions, parameters, and nodes-predictive of AML progression, we present an explainable machine learning algorithm based on high-performance evolutionary computation. The results demonstrate that the predictive methodology could accurately estimate the clinical dynamics of AML progression in terms of blast percentages for both training and novel patients. This study demonstrates that the developed explainable machine learning approach can successfully predict AML progression by leveraging the heterogeneous and longitudinal dynamics of patients' clinical data. More importantly, this methodology shows significant potential for application in modeling the progression dynamics of other acute diseases, providing a flexible and adaptable framework for advancing clinical and translational research.

Ethical implications related to processing of personal data and artificial intelligence in humanitarian crises: a scoping review

BACKGROUND

Humanitarian organizations are rapidly expanding their use of data in the pursuit of operational gains in effectiveness and efficiency. Ethical risks, particularly from artificial intelligence (AI) data processing, are increasingly recognized yet inadequately addressed by current humanitarian data protection guidelines. This study reports on a scoping review that maps the range of ethical issues that have been raised in the academic literature regarding data processing of people affected by humanitarian crises.

METHODS

We systematically searched databases to identify peer-reviewed studies published since 2010. Data and findings were standardized, grouping ethical issues into the value categories of autonomy, beneficence, non-maleficence, and justice. The study protocol followed Arksey and O'Malley's approach and PRISMA reporting guidelines.

RESULTS

We identified 16,200 unique records and retained 218 relevant studies. Nearly one in three (n = 66) discussed technologies related to AI. Seventeen studies included an author from a lower-middle income country while four included an author from a low-income country. We identified 22 ethical issues which were then grouped along the four ethical value categories of autonomy, beneficence, non-maleficence, and justice. Slightly over half of included studies (n = 113) identified ethical issues based on real-world examples. The most-cited ethical issue (n = 134) was a concern for privacy in cases where personal or sensitive data might be inadvertently shared with third parties. Aside from AI, the technologies most frequently discussed in these studies included social media, crowdsourcing, and mapping tools.

CONCLUSIONS

Studies highlight significant concerns that data processing in humanitarian contexts can cause additional harm, may not provide direct benefits, may limit affected populations' autonomy, and can lead to the unfair distribution of scarce resources. The increase in AI tool deployment for humanitarian assistance amplifies these concerns. Urgent development of specific, comprehensive guidelines, training, and auditing methods is required to address these ethical challenges. Moreover, empirical research from low and middle-income countries, disproportionally affected by humanitarian crises, is vital to ensure inclusive and diverse perspectives. This research should focus on the ethical implications of both emerging AI systems, as well as established humanitarian data management practices.

TRIAL REGISTRATION

Not applicable.

Leveraging pretrained language models for seizure frequency extraction from epilepsy evaluation reports

Seizure frequency is essential for evaluating epilepsy treatment, ensuring patient safety, and reducing risk for Sudden Unexpected Death in Epilepsy. As this information is often described in clinical narratives, this study presents an approach to extracting structured seizure frequency details from such unstructured text. We investigated two tasks: (1) extracting phrases describing seizure frequency, and (2) extracting seizure frequency attributes. For both tasks, we fine-tuned three BERT-based models (bert-large-cased, biobert-large-cased, and Bio_ClinicalBERT), as well as three generative large language models (GPT-4, GPT-3.5 Turbo, and Llama-2-70b-hf). The final structured output integrated the results from both tasks. GPT-4 attained the best performance across all tasks with precision, recall, and F1-score of 86.61%, 85.04%, and 85.79% respectively for frequency phrase extraction; 90.23%, 93.51%, and 91.84% for seizure frequency attribute extraction; and 86.64%, 85.06%, and 85.82% for the final structured output. These findings highlight the potential of fine-tuned generative models in extractive tasks from limited text strings.

A comparative analysis of CDC and AI-generated health information using computer-aided text analysis

BACKGROUND

AI-generated content is easy to access. Members of the public use it as an alternative or to supplement official sources, such as the Centers for Disease Control and Prevention (CDC). However, the quality and reliability of AI-generated health information is questionable. This study aims to understand how AI-generated health information differs from that provided by the CDC, particularly in terms of sentiment, readability, and overall quality. Language expectancy theory serves as a framework and offers insights into how people's expectations of message content from different sources can influence perceived credibility and persuasiveness of such information.

METHODS

Computer-aided text analysis was used to analyze 20 text entries from the CDC and 20 entries generated by ChatGPT 3.5. Content analysis utilizing human coders was used to assess the quality of information.

RESULTS

ChatGPT used more negative sentiments, particularly words associated with anger, sadness, and disgust. The CDC's health messages were significantly easier to read than those generated by ChatGPT. Furthermore, ChatGPT's responses required a higher reading grade level. In terms of quality, the CDC's information was a little higher quality than that of ChatGPT, with significant differences in DISCERN scores.

CONCLUSION

Public health professionals need to educate the general public about the complexity and quality of AI-generated health information. Health literacy programs should address topics about quality and readability of AI-generated content. Other recommendations for using AI-generated health information are provided.

Deep learning-based identification of patients at increased risk of cancer using routine laboratory markers

Early screening for cancer has proven to improve the survival rate and spare patients from intensive and costly treatments due to late diagnosis. Cancer screening in the healthy population involves an initial risk stratification step to determine the screening method and frequency, primarily to optimize resource allocation by targeting screening towards individuals who draw most benefit. For most screening programs, age and clinical risk factors such as family history are part of the initial risk stratification algorithm. In this paper, we focus on developing a blood marker-based risk stratification approach, which could be used to identify patients with elevated cancer risk to be encouraged for taking a diagnostic test or participate in a screening program. We demonstrate that the combination of simple, widely available blood tests, such as complete blood count and complete metabolic panel, could potentially be used to identify patients at risk for colorectal, liver, and lung cancers with areas under the ROC curve of 0.76, 0.85, 0.78, respectively. Furthermore, we hypothesize that such an approach could not only be used as pre-screening risk assessment for individuals but also as population health management tool, for example to better interrogate the cancer risk in certain sub-populations.

The European Open Science Cloud as a common good Potentials and limitations of this endeavour

The European Open Science Cloud (EOSC) is envisioned as a transformative platform for advancing Open Science, aimed at benefiting a diverse array of stakeholders, including researchers, innovators, institutions, and the broader public. To fully harness EOSC's potential as a common good, capable of delivering services to the research community such to potentially transform the way scientific production and communication is done, we address critical barriers that may actually restrict the equitable access and the optimal use of such services. In particular, we emphasize that key resources as required to access and exploit EOSC's advanced FAIR-data services - such as data-processing algorithms - are, in fact, intrinsically limited and the access will be competitive. Governance and funding of EOSC present challenges associated with its effective openness in terms of accessibility to resources for its advanced exploitation.

Automatic Feature Selection for Imbalanced Echocardiogram Data Using Event-Based Self-Similarity

Background and Objective: Using echocardiogram data for cardiovascular disease (CVD) can lead to difficulties due to imbalanced datasets, leading to biased predictions. Machine learning models can enhance prognosis accuracy, but their effectiveness is influenced by optimal feature selection and robust classification techniques. This study introduces an event-based self-similarity approach to enhance automatic feature selection approach for imbalanced echocardiogram data. Critical features correlated with disease progression were identified by leveraging self-similarity patterns. This study used an echocardiogram dataset, visual presentations of high-frequency sound wave signals, and data of patients with heart disease who are treated using three treatment methods: catheter ablation, ventricular defibrillator, and drug control-over the course of three years. Methods: The dataset was classified into nine categories and Recursive Feature Elimination (RFE) was applied to identify the most relevant features, reducing model complexity while maintaining diagnostic accuracy. Machine learning classification models, including XGBoost and CATBoost, were trained and evaluated. Results: Both models achieved comparable accuracy values, 84.3% and 88.4%, respectively, under different normalization techniques. To further optimize performance, the models were combined into a voting ensemble, improving feature selection and predictive accuracy. Four essential features-age, aorta (AO), left ventricular (LV), and left atrium (LA)-were identified as critical for prognosis and were found in Random Forest (RF)-voting ensemble classifier. The results underscore the importance of feature selection techniques in handling imbalanced datasets, improving classification robustness, and reducing bias in automated prognosis systems. Conclusions: Our findings highlight the potential of machine learning-driven echocardiogram analysis to enhance patient care by providing accurate, data-driven assessments.

Off-Target Interactions of Vancomycin with Vascular Wall Involving Elastin-Induced Self-Assembly

Off-target effects, which arise from drug interactions in nontarget tissues, can lead to unfavored side effects. The treatment efficacy of vancomycin (Vanco) in Gram-positive bacterial infections is often compromised by the frequent occurrence of Vanco-induced vascular injury. However, the potential targets and underlying molecular mechanisms of this phenomenon remain unclear. Here, we developed multidimensional two-photon imaging for dynamic tracking of fluorescently labeled Vanco in vivo, characterizing the molecular behavior of Vanco in situ after administration and providing the first direct evidence of its interactions with vascular wall. Morphological analysis combined with colocalization imaging identified elastin within the vascular wall as the molecular target. After binding, Vanco underwent self-assembly into forming irregular nanoaggregates, primarily driven by electrostatic and hydrophobic forces. This persistent binding and self-assembly on the elastic lamina resulted in significant endothelial cytotoxicity and subsequent apoptosis, suggesting a mechanistic link to the vascular injury observed in clinical settings. Taken together, our findings revealed off-target molecular interactions between Vanco and vascular elastin in situ, highlighting the importance of considering unintended drug-vascular interactions.

In silico genomic insights into bacteriophages infecting ESBL-producing Escherichia coli from human, animal, and environmental sources

BACKGROUND

The emergence of antimicrobial resistance (AMR) in Escherichia coli, particularly extended-spectrum beta-lactamase-producing E. coli (ESBL-EC), is a global public health concern. Bacteriophages (phages) play a significant role in bacterial evolution and the spread of antibiotic resistance genes (ARGs). This study investigates prophages integrated within ESBL-EC genomes to assess their diversity, gene content, and potential contributions to ESBL-EC persistence across human, animal, and environmental reservoirs. Between May and December 2020, a cross-sectional study was conducted in Abuja and Lagos, collecting 448 stool, cecal, and environmental samples from abattoir workers, slaughtered cattle, and the abattoir environment. ESBL-EC genomes from these samples, obtained in an earlier study, were analyzed for phage regions using PHASTER. Intact prophages were analyzed in silico using computational tools to detect ARGs, ESBL genes, virulence factors, and heavy metal resistance. Their genomic relationships were examined with statistical significance of p < 0.05.

RESULTS

Out of 448 samples, ESBL-EC prevalence was 21.7% (97/448). Among 97 ESBL-EC isolates, 646 prophage regions were detected, with 30% (194/646) classified as intact phages. Among the 158 phages with genus assignments, Punavirus was the most prevalent (60.1%). Escherichia was the most frequent predicted host (308/646), particularly in cattle (n = 143) and human (n = 124) sources. Among ESBL-EC genomes, 83.5% (81/97) with intact phages carried phage-associated ARGs, 76.3% (74/97) carried phage-associated ESBL genes, 18.6% (18/97) harbored phage-associated virulence factors, 15.5% (15/97) contained phage-associated plasmids, and 10.3% (10/97) had heavy metal resistance. The most prevalent phage-associated ARGs detected were qnrS1 (73/81) and blaCTX-M-15 (72/81). Two isolates recovered from abattoir workers carried two phage-like plasmids, each harboring either tet(A) or blaCTX-M-55 gene. The predominant phage lifestyles were temperate (n = 182), mainly in the Peduoviridae family, and lytic (n = 12) in the Punavirus genus.

CONCLUSION

This is the first study in Nigeria to characterize phages in ESBL-EC isolates at the One Health interface. The presence of intact phages in humans, animals, and the environment underscores the complex interactions shaping phage ecology. The discovery of ARGs, virulence genes, and heavy metal resistance within prophages suggests a potential role in AMR dissemination. Future research should focus on elucidating mechanisms of ARG transfer mediated by phages in One Health settings.

Estimating baselines of Raman spectra based on transformer and manually annotated data

Raman spectroscopy is a powerful and non-invasive analytical method for determining the chemical composition and molecular structure of a wide range of materials, including complex biological tissues. However, the captured signals typically suffer from interferences manifested as noise and baseline, which need to be removed for successful data analysis. Effective baseline correction is critical in quantitative analysis, as it may impact peak signature derivation. Current baseline correction methods can be labor-intensive and may require extensive parameter adjustment depending on the input spectrum characteristics. In contrast, deep learning-based baseline correction models trained across various materials, offer a promising and more versatile alternative. This study reports an approach to manually identify the ground-truth baselines for eight different biological materials through extensively tuning the parameters of three classical baseline correction methods, Modified Multi-Polynomial Fit (Modpoly), Improved Modified Multi-Polynomial Fitting (IModpoly), and Adaptive Iteratively Reweighted Penalized Least Squares (airPLS), and combining the outputs to best fit the training data. We designed a one-dimensional Transformer (1dTrans) tailored to fit Raman spectral data for estimating their baselines, and evaluated its performance against convolutional neural network (CNN), ResUNet, and three aforementioned parametric methods. The 1dTrans model achieved lower mean absolute error (MAE) and spectral angle mapper (SAM) scores when compared to the other methods in both development and evaluation of the manually labeled original raw Raman spectra, highlighting the effectiveness of the method in Raman spectra pre-processing.

Presumed periodontitis and multimorbidity patterns: a prospective cohort study in the UK Biobank

OBJECTIVES

To examine the pattern of multimorbidity among people with high risk of periodontitis.

MATERIALS AND METHODS

Over 358,000 UK Biobank participants aged 40-69 years at baseline who took part in the baseline assessment and answered mouth/teeth dental problems were included (2006-2010). Cox proportional hazard models and logistic regression models were used to estimate the association of the risk factors of periodontitis with chronic diseases and multimorbidity, stratified by follow-up time.

RESULTS

A total of 154,985 participants developed multimorbidity during follow-up. We observed increased risk of multimorbidity among participants with presumed periodontitis (adjusted HR = 1.06, 95% confidence interval [CI] = 1.05-1.08), especially in those participants with age < 50 years old (adjusted HR = 1.11, 95% CI = 1.08-1.14). Among the different multimorbidity patterns, presumed periodontitis was mainly associated with the mental disorder pattern and metabolic and vascular disease pattern.

CONCLUSIONS

Presumed periodontitis was positively associated with multimorbidity, even more so in younger age. We need to pay more attention to the prevention of periodontitis in the early stage to reduce the burden of multimorbidity in the future.

CLINICAL RELEVANCE

Early life interventions to prevent periodontitis are crucial to reduce the incidence of multimorbidity and enhance the quality of life in older adults. Additionally, greater attention should be given to the mental and cardiovascular metabolic health of patients with periodontitis.

Dementia classification using two-channel electroencephalography features

This study aimed to develop a novel classification model using wearable two-channel electroencephalography (EEG) data to differentiate between patients with dementia and normal controls (NCs). We employed an extreme gradient boosting (Xgboost) model combined with recursive feature elimination with cross-validation (RFECV) to classify patients and NCs. The study included 54 NCs and 29 patients with dementia. Resting-state EEG was recorded, and Mini-Mental Status Exam (MMSE) and Clinical Dementia Rating (CDR) assessments were conducted. Significant differences were observed in peak frequency (PF), alpha (A), theta (T), the ratio of alpha to theta (A/T), the ratio of alpha to low-beta (A/BL), and coherence (CH) between patients and NCs. Patients with dementia exhibited decreases in PF, CH_A/T, CH_A/BL, A/T, and A/BL, while an increase in T was noted. The primary finding was that the Xgboost model, a tree ensemble classification, achieved a balanced accuracy of 97.05% with the RFECV-selected feature, which was PF. This study suggests that the novel Xgboost with RFECV classification model using two-channel EEG data could be a valuable tool for diagnosing dementia.

The application of machine learning models in a resource-constrained environment

BACKGROUND

Machine learning models (MLMs) used to influence surgical decision making often require large and complex datasets upon which to train. However, there is a paucity of literature pertaining to the ability to apply standard MLMs to small ICU datasets within resource-constrained institutions.

METHODS

ML models were applied to a prospective cohort of critically ill mechanically ventilated patients from a teaching hospital in rural Kenya. Characteristics included an ICU scoring system specifically for resource-constrained environments (Tropical Intensive Care Score (TropICS)). Outputs included AUC of the ROC and the feature importance table. Python-based MLMs included XGBoost and KNN. AUC of the ROC was calculated to predict mortality as the primary endpoint.

RESULTS

There were 294 patients, with an average age of 40.2 years, 64.3% male, 23.8% trauma, and an overall mortality of 60.2%. With respect to mortality patients who died were older (43.5 versus 35 years; p < 0.001), but with no difference in male gender (64.8% versus 63.8%; p = 0.9), or having been transferred from outside facilities (34% versus 21.5%; p = 0.5). Whilst there was no difference in the rate of tachycardia or acidosis, patients who died were more likely to present with hemodynamic instability (31% versus 6%; p < 0.001) and higher clinical severity scores. In predicting mortality, the ML models performed very well (XGBoost AUC = 0.82). Within MLM feature importance, the Tropical Intensive Care Score (TropICS) performed as well as APACHE-II and the SAPS.

CONCLUSION

ML models can be effectively applied to a small ICU dataset within resource-constrained environments. ML models must demonstrate functionality prior to incorporating within prospective clinical predictive models. Contextualized ICU scoring systems (TropICS) performed well within MLMs.

Agri-food data spaces: Highlighting the need for a farm-centered strategy

This paper explores the potential of digitalisation in agriculture to improve the sustainability of agriculture production and industrial sectors, contributing to the twin digital and green transition. These systems can facilitate and enhance competitiveness by leveraging on mutually reinforcing transformations. The European Commission has proposed the creation of Common European Data Spaces in specific sectors to support such a transition. We focus on the agri-food domain, considering farmers and other actors in the food chain. The aim is to identify needs, priorities, opportunities, and barriers to a Common European Data Space for agriculture and food systems, thus going beyond the sectoral European Data Space for agriculture already under current development. In addition, this work looks at strategies for introducing the aforementioned novel data space and evidence of benefits for farmers, who are a key component of agricultural and food systems. To accomplish this, the concept of data spaces is presented, analysing main components, functions, and potential challenges and opportunities for data sharing and reuse, with the agri-food context as the main focus. It also presents current and future scenarios for data use at different decision-making levels, focusing on the specific role of farmers in the digital ecosystem. Additionally, it outlines the basic principles for an inclusive agri-food data strategy.

Anxiety about artificial intelligence from patient and doctor-physician

OBJECTIVE

This paper investigates the anxiety surrounding the integration of artificial intelligence (AI) in doctor-patient interactions, analyzing the perspectives of both patients and healthcare providers to identify key concerns and potential solutions.

METHODS

The study employs a comprehensive literature review, examining existing research on AI in healthcare, and synthesizes findings from various surveys and studies that explore the attitudes of patients and doctors towards AI applications in medical settings.

RESULTS

The analysis reveals that patient anxiety encompasses algorithm aversion, robophobia, lack of humanistic care, challenges in human-machine interaction, and concerns about AI's universal applicability. Doctors' anxieties stem from fears of replacement, legal liabilities, emotional impacts of work environment changes, and technological apprehension. The paper highlights the need for patient participation, humanistic care, improved interaction methods, educational training, and policy guidelines to foster public understanding and trust in AI.

CONCLUSION

The paper concludes that addressing AI anxiety in doctor-patient relationships is crucial for successfully integrating AI in healthcare. It emphasizes the importance of respecting patient autonomy, addressing the lack of humanistic care, and improving patient-AI interaction to enhance the patient experience and reduce medical errors.

PRACTICE IMPLICATIONS

The study suggests that future research should focus on understanding the needs and concerns of patients and doctors, strengthening medical humanities education, and establishing policies to guide the ethical use of AI in medicine. It also recommends public education to enhance understanding and trust in AI to improve medical services and ensure professional development and stable work environment for doctors.

Lightweight and hybrid transformer-based solution for quick and reliable deepfake detection

Introduction

Rapid advancements in artificial intelligence and generative artificial intelligence have enabled the creation of fake images and videos that appear highly realistic. According to a report published in 2022, approximately 71% of people rely on fake videos and become victims of blackmail. Moreover, these fake videos and images are used to tarnish the reputation of popular public figures. This has increased the demand for deepfake detection techniques. The accuracy of the techniques proposed in the literature so far varies with changes in fake content generation techniques. Additionally, these techniques are computationally intensive. The techniques discussed in the literature are based on convolutional neural networks, Linformer models, or transformer models for deepfake detection, each with its advantages and disadvantages.

Methods

In this manuscript, a hybrid architecture combining transformer and Linformer models is proposed for deepfake detection. This architecture converts an image into patches and performs position encoding to retain spatial relationships between patches. Its encoder captures the contextual information from the input patches, and Gaussian Error Linear Unit resolves the vanishing gradient problem.

Results

The Linformer component reduces the size of the attention matrix. Thus, it reduces the execution time to half without compromising accuracy. Moreover, it utilizes the unique features of transformer and Linformer models to enhance the robustness and generalization of deepfake detection techniques. The low computational requirement and high accuracy of 98.9% increase the real-time applicability of the model, preventing blackmail and other losses to the public.

Discussion

The proposed hybrid model utilizes the strength of the transformer model in capturing complex patterns in data. It uses the self-attention potential of the Linformer model and reduces the computation time without compromising the accuracy. Moreover, the models were implemented on patch sizes of 6 and 11. It is evident from the obtained results that increasing the patch size improves the performance of the model. This allows the model to capture fine-grained features and learn more effectively from the same set of videos. The larger patch size also enables the model to better preserve spatial details, which contributes to improved feature extraction.

Utilizing large language models for gastroenterology research: a conceptual framework

Large language models (LLMs) transform healthcare by assisting clinicians with decision-making, research, and patient management. In gastroenterology, LLMs have shown potential in clinical decision support, data extraction, and patient education. However, challenges such as bias, hallucinations, integration with clinical workflows, and regulatory compliance must be addressed for safe and effective implementation. This manuscript presents a structured framework for integrating LLMs into gastroenterology, using Hepatitis C treatment as a real-world application. The framework outlines key steps to ensure accuracy, safety, and clinical relevance while mitigating risks associated with artificial intelligence (AI)-driven healthcare tools. The framework includes defining clinical goals, assembling a multidisciplinary team, data collection and preparation, model selection, fine-tuning, calibration, hallucination mitigation, user interface development, integration with electronic health records, real-world validation, and continuous improvement. Retrieval-augmented generation and fine-tuning approaches are evaluated for optimizing model adaptability. Bias detection, reinforcement learning from human feedback, and structured prompt engineering are incorporated to enhance reliability. Ethical and regulatory considerations, including the Health Insurance Portability and Accountability Act, General Data Protection Regulation, and AI-specific guidelines (DECIDE-AI, SPIRIT-AI, CONSORT-AI), are addressed to ensure responsible AI deployment. LLMs have the potential to enhance decision-making, research efficiency, and patient care in gastroenterology, but responsible deployment requires bias mitigation, transparency, and ongoing validation. Future research should focus on multi-institutional validation and AI-assisted clinical trials to establish LLMs as reliable tools in gastroenterology.

Challenges and applications in generative AI for clinical tabular data in physiology

Recent advancements in generative approaches in AI have opened up the prospect of synthetic tabular clinical data generation. From filling in missing values in real-world data, these approaches have now advanced to creating complex multi-tables. This review explores the development of techniques capable of synthesizing patient data and modeling multiple tables. We highlight the challenges and opportunities of these methods for analyzing patient data in physiology. Additionally, it discusses the challenges and potential of these approaches in improving clinical research, personalized medicine, and healthcare policy. The integration of these generative models into physiological settings may represent both a theoretical advancement and a practical tool that has the potential to improve mechanistic understanding and patient care. By providing a reliable source of synthetic data, these models can also help mitigate privacy concerns and facilitate large-scale data sharing.

Machine Learning Unveils Sphingolipid Metabolism's Role in Tumour Microenvironment and Immunotherapy in Lung Cancer

TME is a core player in the development of a cancerous lesion, the immune evasive potential of the lesion, and its response to therapy. Sphingolipid metabolism, which governs a number of cellular processes, has been recognised as a player involved in the control of immune heterogeneity within the TME. Sphingolipid metabolism-related genes prevalent in the TME of LUAD and LUSC were identified using transcriptomic analysis and clinical samples from the TCGA and GTEx databases. Lasso regression and survival SVM in the Etra Application were employed as machine learning algorithms to determine patient outcomes and to reveal key immune factors associated with gene expression and chemotherapeutic response. Gene expression in lung cancer cells was explored through scRNA-seq data. Thereafter, mediation impact analysis was further performed to explain the defined relation between the immune cell subsets and sphingolipid metabolites and their risk impact on lung cancers. Genes involved in sphingolipid metabolism were dysregulated in lung cancer, correlating with immune cell infiltration and TME remodelling. Lasso regression identified ASAH1 and SMPD1 as strong prognostic markers. scRNA-seq revealed higher gene expression in T cells, macrophages and fibroblasts. Sphingomyelin partially mediated the link between T lymphocyte abundance and lung cancer risk. High-risk phenotypes exhibited enhanced immune evasion via altered regulatory T cell and macrophage polarisation. This research highlights the contribution of sphingolipid metabolism in shaping the TME and its implications for immunotherapy.

Methods for estimating resting energy expenditure in intensive care patients: A comparative study of predictive equations with machine learning and deep learning approaches

BACKGROUND

Accurate estimation of resting energy expenditure (REE) is critical for guiding nutritional therapy in critically ill patients. While indirect calorimetry (IC) is the gold standard for REE measurement, it is not routinely feasible in clinical settings due to its complexity and cost. Predictive equations (PEs) offer a simpler alternative but are often inaccurate in critically ill populations. While recent advancements in machine learning (ML) and deep learning (DL) offer potential for improving REE estimation by capturing complex relationships between physiological variables, these approaches have not yet been widely applied or validated in critically ill populations.

METHODOLOGY

This prospective study compared the performance of nine commonly used PEs, including the Harris-Benedict (H-B1919), Penn State, and TAH equations, with ML models (XGBoost, Random Forest Regressor [RFR], Support Vector Regression), and DL models (Convolutional Neural Networks [CNN]) in estimating REE in critically ill patients. A dataset of 300 IC measurements from an intensive care unit (ICU) was used, with REE measured by both IC and PEs. The ML/DL models were trained using a combination of static (i.e., age, height, body weight) and dynamic (i.e., minute ventilation, body temperature) variables. A five-fold cross validation was performed to assess the model prediction performance using the root mean square error (RMSE) metric.

RESULTS

Of the PEs analysed, H-B1919 yielded the lowest RMSE at 362 calories. However, the XGBoost and RFR models significantly outperformed all PEs, achieving RMSE values of 199 and 200 calories, respectively. The CNN model demonstrated the poorest performance among ML models, with an RMSE of 250 calories. The inclusion of additional categorical variables such as body mass index (BMI) and body temperature classes slightly reduced RMSE across ML and DL models. Despite data augmentation and imputation techniques, no significant improvements in model performance were observed.

CONCLUSION

ML models, particularly XGBoost and RFR, provide more accurate REE estimations than traditional PEs, highlighting their potential to better capture the complex, non-linear relationships between physiological variables and REE. These models offer a promising alternative for guiding nutritional therapy in clinical settings, though further validation on independent datasets and across diverse patient populations is warranted.

Combining machine learning models and rule engines in clinical decision systems: Exploring optimal aggregation methods for vaccine hesitancy prediction

BACKGROUND

With the increasing application of artificial intelligence (AI) technologies in the healthcare sector and the emergence of new solutions, such as large language models, there is a growing need to combine medical knowledge, often expressed as clinical rules, with advances in machine learning (ML) offering higher prediction accuracy at the expense of decision-making transparency.

PURPOSE

This study investigates the efficacy of various aggregation methods combining the decisions of an AI model and a clinical rule-based (RB) engine in predicting vaccine hesitancy to maximize the effectiveness of patient incentive programs. This is the first study of parallel ensemble of rules and machine learning in clinical context proposing RB confidence-led fusion of ML and RB inference.

METHODS

A clinical decision system for predicting hesitation to vaccinate is developed based on a differentially private set of longitudinal health records of 974,000 US patients and clinical rules obtained from the present literature. Various approaches based on possibility theory have been explored to maximize classification accuracy, capture and hurdle rates while ensuring trustworthiness in clinical interventions.

RESULTS

Our findings reveal that the hybrid approach outperforms the individual models and RB systems when transparency and accuracy are critical. A RB confidence-led approach emerged as the most effective method. The aggregation of mismatched classes relies on RB results when the RB engine has high confidence (expressed as more than the median degree of membership to the vaccination hesitation output function) and on ML predictions when the RB engine exhibits lower confidence.

CONCLUSIONS

Implementing such an aggregation method preserves the accuracy and capture rates of a clinical decision system, while potentially improving acceptance among healthcare providers.

Neurophysiological data augmentation for EEG-fNIRS multimodal features based on a denoising diffusion probabilistic model

BACKGROUND AND OBJECTIVE

The hybrid brain computer interfaces (BCI) combining electroencephalogram (EEG) and functional near-infrared spectroscopy (fNIRS) have attracted extensive attention for overcoming the decoding limitations of the single-modality BCI. With the deepening application of deep learning approaches in BCI systems, its significant performance improvement has become apparent. However, the scarcity of brain signal data limits the performance of deep learning models.

METHODS

In this study, we proposed an EEG-fNIRS data augmentation framework based on the combination of denoising diffusion probabilistic model (DDPM) and adding Gaussian noise (EFDA-CDG), for enhancing the performance of hybrid BCI systems. Firstly, we unified the temporal and spatial dimensions of EEG and fNIRS by manually extracting features and spatial mapping interpolation to create EEG-fNIRS joint distribution samples. Then, the DDPM generative model was combined with the traditional method of adding Gaussian noise to provide richer training data for the classifier. Finally, we constructed a classification module that applies EEG feature attention and fNIRS terrain attention to improve classification accuracy.

RESULTS

In order to evaluate the effectiveness of EFDA-CDG framework, experiments were conducted and fully validated on three publicly available databases and one self-collected database. In the context of a participant-dependent training approach, our method achieves accuracy rates of 82.02% for motor imagery, 91.93% for mental arithmetic, and 90.54% for n-back tasks on public databases. Additionally, our method boasts an accuracy rate of 97.82% for drug addiction discrimination task on the self-collected database.

CONCLUSIONS

EFDA-CDG framework successfully facilitates data augmentation, thereby enhancing the performance of EEG-fNIRS hybrid BCI systems.

Blowfly genomics: current insights, knowledge gaps, and future perspectives

Blowflies (Calliphoridae) form a diverse, species-rich group, yet publicly available genome assemblies are limited to only 16 species, despite recent genomic advances. This knowledge gap extends to mitogenomes and barcode databases, which mainly focus on medically and veterinary-important species. While blowfly phylogenetics has progressed, additional genome sequencing is crucial for various subfamilies, given their diverse life histories. This review presents a quantitative overview of available genetic information for blowflies, highlighting substantial gaps in public databases. DNA barcodes, mitogenomes, and genomes represent only 16.5% (342 species), ∼3% (53 species), and <1% (16 species) of known family diversity, respectively. While 183 genomics-related calliphorid BioProjects are recorded by NCBI, many subfamilies and genera have limited or no genomic representation, impacting studies on identification, systematics, phylogenetics, and evolution. We stress the urgent need for high-quality reference genomes and highlight target species representing all blowfly subfamilies to support a new era of rapid, low-cost genomic research.

The COVID-19 pandemic and the worldwide online interest in telepsychiatry: an infodemiological study from 2004 to 2022

Introduction

Studies call for the further assessment and understanding of public interests and concerns about telepsychiatry, especially during the COVID-19 pandemic. Since telepsychiatry services are accessed through the Internet, this study analyzed online searches and queries to determine telepsychiatry-related interests and concerns over time. The findings can inform the development and customization of online telepsychiatry resources and services, enabling a more effective response to public needs.

Materials and methods

This study determined public concerns and interests in telepsychiatry using data from Google Trends and Wikipedia from 2004 to 2022. These platforms were selected for their large global market share. After describing the data, bootstrap for independent sample tests of search volumes and Wikipedia page views before and during the COVID-19 pandemic.

Results

The highest interest in telepsychiatry was observed in high-income countries. Search volumes for telepsychiatry increased, while Wikipedia page views decreased during the COVID-19 pandemic. The top and rising queries that can be incorporated into telepsychiatry websites include telepsychiatry concepts, jobs, services, costs, and locations.

Discussion

The findings support that the use of the Internet for telepsychiatry information increased compared to previous years, especially during the start of the COVID-19 pandemic. There may also be a higher interest in telepsychiatry among high-income nations compared to low and middle-income countries. Furthermore, the study also supports that digital information should be tailored to respond to public needs and expectations by incorporating telepsychiatry-related concepts, jobs, services, costs, and locations.

A Survey of Perspectives and Educational Needs of Canadian Oncology Residents on Artificial Intelligence

This study evaluated the perspectives and educational needs of Canadian oncology residents with regard to artificial intelligence (AI) in medicine, exploring the influence of factors such as program of choice, gender, and tech literacy on their attitudes towards AI. An ethics-approved survey collected anonymous responses from Canadian oncology residents from December 2022 to July 2023. Comparisons by demographics were made using Chi-square and Mann-Whitney U tests. A total of 57 residents and fellows responded out of an expected 182, with representation from each oncology training program in Canada. Over half of the participants were male (63.2%), with radiation oncology programs being better represented than medical oncology programs (68.4% vs. 31.6%). There was balanced representation across all years of training. Most trainees (73%) were interested in learning more about AI, and many believed the topic should be formally taught during residency (63%), preferably through workshops (79%). Among evaluated factors, tech literacy showed the most impact over AI perspectives, driving a perception shift towards viewing AI as an improvement tool, rather than as a threat to professionals. In conclusion, Canadian oncology residents anticipate AI's growing influence in medicine but face educational deficiencies. Gender, oncology discipline, and self-reported tech literacy impact attitudes toward AI, highlighting the need for inclusive education.

The ethics of artificial intelligence systems in healthcare and medicine: from a local to a global perspective, and back

Artificial intelligence systems (ai-systems) (e.g. machine learning, generative artificial intelligence), in healthcare and medicine, have been received with hopes of better care quality, more efficiency, lower care costs, etc. Simultaneously, these systems have been met with reservations regarding their impacts on stakeholders' privacy, on changing power dynamics, on systemic biases, etc. Fortunately, healthcare and medicine have been guided by a multitude of ethical principles, frameworks, or approaches, which also guide the use of ai-systems in healthcare and medicine, in one form or another. Nevertheless, in this article, I argue that most of these approaches are inspired by a local isolationist view on ai-systems, here exemplified by the principlist approach. Despite positive contributions to laying out the ethical landscape of ai-systems in healthcare and medicine, such ethics approaches are too focused on a specific local healthcare and medical setting, be it a particular care relationship, a particular care organisation, or a particular society or region. By doing so, they lose sight of the global impacts ai-systems have, especially environmental impacts and related social impacts, such as increased health risks. To meet this gap, this article presents a global approach to the ethics of ai-systems in healthcare and medicine which consists of five levels of ethical impacts and analysis: individual-relational, organisational, societal, global, and historical. As such, this global approach incorporates the local isolationist view by integrating it in a wider landscape of ethical consideration so to ensure ai-systems meet the needs of everyone everywhere.

The impact of maternal depression and anxiety on well-baby visit attendance: a retrospective cohort study of 6,699 PRAMS participants from 2016-2021

PURPOSE

The objective of this study was to determine the independent effects of maternal mental health difficulties (MHD) during the preconception, prenatal, and postnatal periods on well-baby visit (WBV) attendance in a population-wide sample through retrospective analysis of Pregnancy Risk Assessment Monitoring System (PRAMS) data.

METHODS

This secondary analysis utilized data from the 2016 - 2021 New Jersey PRAMS, yielding 6,699 participants (mean age = 30.8 years). Survey-weighted means, confidence intervals, and percentages were used to describe sociodemographic, mental health, and WBV variables across all participants. Logistic regression with complex survey weights and multiple imputation of missing data was implemented to test associations between sociodemographic factors, maternal MHDs, and WBV attendance.

RESULTS

The weighted prevalence of missing the 1-week checkup or having never attended a WBV during the first six months postpartum was 4.3% (95% CI: 3.8% - 5.0%; n = 260) and 1.4% (1.1% - 2.0%; n = 98), respectively. Preconception depression (n = 553; 7.7%, 7.0% - 8.0%), prenatal depression (n = 481; 6.5%, 5.9% - 7.0%;), preconception anxiety (n = 1,007; 15.2%, 14.2% - 16.0%), and prenatal anxiety (n = 570; 8.44%, 7.7% - 9.0%) were not associated with 1-week checkup attendance. However, women with preconception depression were more than twice as likely to have never attended a WBV (OR = 2.43, 1.01 - 5.82). Multiple social determinants and demographic variables were associated with greater odds of missing WBVs, including middle household income, receiving government-issued health insurance or being uninsured, Hispanic ethnicity, and Spanish as a primary language.

CONCLUSIONS

Preconception depression, middle household income, receiving government-issued health insurance, being uninsured, Hispanic ethnicity, and Spanish as a primary language may decrease attendance of WBVs, and the mediating role of preconception depression in infant health outcomes warrants further investigation.