Category: Nevin Manimala

Sci Rep. 2025 Sep 26;15(1):33214. doi: 10.1038/s41598-025-18306-1.

ABSTRACT

This single-center retrospective study evaluated the performance of four multimodal large language models (MLLMs) (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, Perplexity Sonar Large) in detecting and grading the severity of age-related macular degeneration (AMD) from ultrawide field fundus images. Images from 76 patients (136 eyes; mean age 81.1 years; 69.7% female) seen at the University of California San Diego were graded independently for AMD severity by two junior retinal specialists (and an adjudicating senior retina specialist for disagreements) using the Age-Related Eye Disease Study (AREDS) classification. The cohort included 17 (12.5%) eyes with ‘No AMD’, 18 (13.2%) with ‘Early AMD’, 50 (36.8%) with ‘Intermediate AMD’, and 51 (37.5%) with ‘Advanced AMD’. Between December 2024 and February 2025, each MLLM was prompted with single images and standardized queries to assess the primary outcomes of accuracy, sensitivity, and specificity in binary disease classification, disease severity grading, open-ended diagnosis, and multiple-choice diagnosis (with distractor diseases). Secondary outcomes included precision, F1 scores, Cohen’s kappa, model performance comparisons, and error analysis. ChatGPT-4o demonstrated the highest accuracy for binary disease classification [mean 0.824 (95% confidence interval (CI)): 0.743, 0.875)], followed by Perplexity Sonar Large [mean 0.815 (95% CI: 0.744, 0.879)], both of which were significantly more accurate (P < 0.00033) Than Gemini 1.5 Pro [mean 0.669 (95% CI: 0.581, 0.743)] and Claude 3.5 Sonnet [mean 0.301 (95% CI: 0.221, 0.375)]. For severity grading, Perplexity Sonar Large was most accurate [mean 0.463 (95% CI: 0.368, 0.537)], though differences among models were not statistically significant. ChatGPT-4o led in open-ended and multiple-choice diagnostic tasks. In summary, while MLLMs show promise for automated AMD detection and grading from fundus images, their current reliability is insufficient for clinical application, highlighting the need for further model development and validation.

PMID:41006661 | DOI:10.1038/s41598-025-18306-1

Commun Med (Lond). 2025 Sep 26;5(1):400. doi: 10.1038/s43856-025-01104-1.

ABSTRACT

BACKGROUND: Knee osteoarthritis (KOA) is common in older adults and may relate to cognitive decline. We explore whether changes in specific brain areas and body inflammation levels help explain this connection, focusing on the hippocampus-a memory-critical brain region.

METHODS: We studied 36 older adults with KOA over time. Using brain scans, we measured volumes of hippocampal subregions (especially the fimbria). Blood tests tracked six inflammation markers, including brain-derived neurotrophic factor (BDNF), interferon-gamma (IFN-γ), programmed death 1(PD-1), recombinant cannabinoid receptor 1 (CNR1), recombinant cannabinoid receptor 2 (CNR2), and T cell immunoglobulin domain and mucin domain 3 (TIM3). Memory was tested using the Wechsler Memory Scale – Chinese Revision (WMS-CR), while global cognition used the Montreal Cognitive Assessment (MoCA). Relationships between knee pain, brain structure, inflammation, and cognition were analyzed statistically.

RESULTS: Here, we show that shrinking fimbria volume predicts cognitive decline in those developing dementia. We identify a robust correlation between fimbria volume and cognitive performance. Higher IFN-γ levels are protective against cognitive decline. Critically, fimbria volume serves as a mediator in the relationship between pain, TIM3/IFN-γ levels, and cognitive scores.

CONCLUSIONS: Fimbria serves as a key pathway through which KOA may drive cognitive impairment, while IFN-γ could help protect memory. Monitoring these hippocampal changes and inflammation levels might help identify at-risk patients early.

PMID:41006612 | DOI:10.1038/s43856-025-01104-1

Pediatr Cardiol. 2025 Sep 26. doi: 10.1007/s00246-025-04044-8. Online ahead of print.

ABSTRACT

Neonatal portal vein thrombosis (PVT) is an uncommon but clinically important vascular complication, often associated with umbilical venous catheter (UVC) use. The optimal management strategy, including the role of anticoagulation, remains uncertain. This retrospective case series included 13 neonates with PVT diagnosed in a level III NICU between January 2021 and December 2024. Clinical and imaging data were compared between infants with spontaneous thrombus resolution and those with persistent thrombosis. The median gestational age was 30.1 weeks, and the median birth weight was 875 g. All infants had UVC placement; every thrombus involved the left portal vein with intrahepatic extension, and all extended into the IVC. Two neonates (15.4%) received anticoagulation; the remainder were managed conservatively. Spontaneous resolution occurred in 6 of 13 cases (46.2%). Earlier diagnosis and higher birth weight were more frequent in the resolution group, although not statistically significant. No thrombus-related acute complications occurred during a median follow-up of 6 months. In this case series, nearly half of neonates with non-occlusive PVT showed spontaneous resolution without anticoagulation. These findings suggest that conservative management can be considered in clinically stable infants, but the short follow-up precludes firm conclusions regarding long-term safety. Ongoing surveillance is essential to detect late complications such as portal hypertension or portosystemic shunting.

PMID:41006583 | DOI:10.1007/s00246-025-04044-8

Sci Rep. 2025 Sep 26;15(1):33035. doi: 10.1038/s41598-025-18079-7.

ABSTRACT

When predicting adverse complications due to Type 2 Diabetes, often two different approaches are taken: predictions based on clinical data or those using administrative health data. No studies have assessed whether these two approaches reach comparable predictions. This study compares the predictive performance of these two data sources and examines the algorithmic fairness of the developed models. We developed XGBoost models to predict the two-year risk of nephropathy, tissue infection, and cardiovascular events in Type 2 Diabetes patients. The models using only clinical data achieved an average AUC of 0.78, while the models using administrative health data alone achieved 0.77. A hybrid model combining both data types resulted in an average AUC of 0.80, across complications. The models showed that laboratory data were key for predicting nephropathy, whereas comorbidity and diabetes age were most important for tissue infection. For cardiovascular events, age and a history of congestive heart failure was the most important predictors. Our analysis identified bias on the feature sex in all three outcomes: models tended to underestimate risk for females and overestimate it for males, indicating a need to address fairness in these applications. This study demonstrates the effectiveness of ML models using both data types for predicting diabetes complications. However, the presence of sex bias highlights the importance of improving model fairness for reliable clinical use.

PMID:41006578 | DOI:10.1038/s41598-025-18079-7

Sci Rep. 2025 Sep 26;15(1):33002. doi: 10.1038/s41598-025-18230-4.

ABSTRACT

Accurately predicting the discharge coefficient (C_d) in weir structures is crucial for improving hydraulic designs and ensuring their safe operation. This study focuses on developing and testing advanced Machine Learning (ML) models to estimate C_d in Semicircular Labyrinth Weirs (SCLWs). The models explored include a Tabular Neural Network (TabNet) optimized with the Moth Flame Optimization algorithm (TabNet-MFO), an Extreme Learning Machine (ELM) enhanced with the Jaya and Firefly Algorithms (ELM-JFO), a Decision Tree (DT), and a Light Gradient Boosting Machine (LightGBM). One of the key innovations in this study is the introduction of the TabNet-MFO framework. Through sensitivity analysis, using tools like the Explainable Boosting Machine (EBM) and SHapley Additive exPlanations (SHAP), the study found that the ratio of upstream flow depth to weir height (h/P) is the most significant factor affecting C_d predictions. Other important factors include the number of weir cycles (N) and the ratio of crest length to weir height (l_C/P). The dataset was split into 75% for train and 25% for validation. The performance of each model was gauged using a number of statistical indicators. They were the coefficient of determination (R²), the Root Mean Square Error (RMSE), the symmetric Mean Absolute Percentage Error (sMAPE), the Scatter Index (SI), and the Weighted Mean Absolute Percentage Error, or WMAPE and along with Taylor diagrams and the Performance Index (PI) for comparison. In the training phase, the ELM-JFO model delivered the best results in predicting C_d, with a PI of 166 and a normalized centered RMSE (E’) of 0.0052. The TabNet-MFO model also performed well, with a PI of 142 and an E’ of 0.0068. The LightGBM and DT models produced good results as well, with PIs of 89.45 and 89.36, respectively. In the testing phase, the TabNet-MFO model remained the top performer (PI = 81.92, E’ = 0.0118), followed by ELM-JFO (PI = 69.71, E’ = 0.0139). LightGBM and DT showed lower accuracy, with PIs of 60.62 and 47.55 and E’ values of 0.0159 and 0.0199, respectively. The novelty of this research lies in combining interpretable and hybrid ML techniques for C_d estimation, offering a reliable alternative to traditional empirical and regression-based methods. These results show the potential of ML in improving flow prediction accuracy and supporting better hydraulic structure design.

PMID:41006552 | DOI:10.1038/s41598-025-18230-4

Sci Rep. 2025 Sep 26;15(1):33092. doi: 10.1038/s41598-025-16712-z.

ABSTRACT

With the promotion of AI applications, people have undergone various changes, especially high school students who have a strong sense of engagement in AI-related activities. In order to explore the interactive mechanism and influencing factors between AI literacy and computing thinking level of high school students in H city, this study focuses on the relationship between student background, AI literacy, and computational thinking. By conducting a questionnaire survey of hundreds of high school students and using SPSS for descriptive statistics, analysis of variance, and correlation analysis, the focus is on exploring the impact and interrelationships between the three dimensions of AI literacy and the five dimensions of computational thinking. A structural equation model (SEM) was constructed by using AMOS software to further explore its internal complex correlation relationship. The results show that parental education and daily use of AI tools significantly affect students’ AI knowledge and skills, while factors such as gender and family location have different degrees of positive or negative effects on creativity, algorithmic thinking, and critical thinking. In addition, artificial intelligence literacy is moderately positively correlated with some dimensions of computational thinking. This study provides empirical support for the rational planning of AI courses in the basic education stage, strengthening the cultivation of students’ computational thinking and optimizing teaching practice.

PMID:41006537 | DOI:10.1038/s41598-025-16712-z

Sci Rep. 2025 Sep 26;15(1):33277. doi: 10.1038/s41598-025-18925-8.

ABSTRACT

Protein glycation is crucial in the pathogenesis of diabetes and its cardiovascular complications. Little is known about the antiglycation properties of amlodipine, a long-acting calcium channel blocker used to treat high blood pressure. In our study, amlodipine’s antiglycoxidant activity was assayed in sugars (glucose, fructose, and ribose), aldehydes (glyoxal), and chloramine T-modified bovine serum albumin (BSA). Aminoguanidine and N-acetylcysteine were used as standard glycation/oxidation inhibitors. The content of oxidation, glycoxidation, and glycation protein products was measured colorimetrically and fluorimetrically. A one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used for statistical analysis. The mechanism of amlodipine’s antiglycoxidant activity was also evaluated using in-silico molecular docking. Amlodipine protects against BSA oxidation, as evidenced by enhanced total thiol content and mitigated protein carbonyls/advanced oxidation protein products. Amlodipine also increased the fluorescence of tryptophan and decreased the contents of kynurenine, N-formylkynurenine, and dityrosine. In addition, amlodipine effectively prevents protein glycation, as evidenced by a reduction in amyloid-beta structure, Amadori products, and advanced glycation end products (AGEs). In in silico analysis, amlodipine’s antiglycation properties were indicated during its interaction with BSA, glycosidases, and AGEs/receptor for AGEs (RAGE) pathway proteins. Among all proteins, amlodipine docked best with c-Jun N-terminal kinases. Summarizing, we have demonstrated the anti-glycation and antioxidant activity of amlodipine in vitro. This effect may be particularly important in patients with diabetes and atherosclerosis, where excessive glycation accelerates the development of vascular complications. Further studies are needed to confirm the antidiabetic activity of amlodipine in vivo.

PMID:41006534 | DOI:10.1038/s41598-025-18925-8

Sci Rep. 2025 Sep 26;15(1):32972. doi: 10.1038/s41598-025-17303-8.

ABSTRACT

Brain metastases occur in 20-30% of patients with malignant tumors. While stereotactic radiosurgery (SRS) offers an effective treatment option for limited brain metastases, managing symptomatic patients with multiple metastases remains challenging due to prolonged treatment times and potential neurological complications. The dynamic conformal arc (DCA) SRS technique using a single-isocenter multi-target (DCA-SIMT) approach facilitates efficient treatment of numerous lesions. This study evaluates survival outcomes and symptom control in patients treated with DCA-SIMT SRS for multiple brain metastases. A registry based analysis was conducted on 123 patients with 560 metastatic CNS lesions treated with DCA-SIMT SRS at the Franciszek Lukaszczyk Oncology Center, Poland, between August 2018 and September 2020. The median survival time was 7.17 months. Patients were assessed for survival, local control, and treatment-related symptoms. Statistical analyses included Cox regression and Kaplan-Meier survival analysis. The 6-month and 12-month survival rates were 57% and 29%, respectively. Local control was achieved in 93% of lesions. The total planning target volume (PTV) was a significant prognostic factor (p = 0.008), with an increase in PTV associated with decreased survival. Patients with PTV ≤ 10 cm³ had significantly longer survival (p = 0.007). Histopathological subtype also influenced outcomes, with sqamous non-small cell lung cancer associated with poorer survival (p = 0.003). Neurological symptoms stabilized or improved in 61% of patients post-treatment, despite a median global V12 of 11.6 cm³, which was not associated with increased toxicity. DCA-SIMT SRS is a viable option for symptomatic patients with multiple brain metastases, offering rapid, precise treatment with significant clinical benefits.

PMID:41006529 | DOI:10.1038/s41598-025-17303-8

Sci Rep. 2025 Sep 26;15(1):33216. doi: 10.1038/s41598-025-17866-6.

ABSTRACT

The SPUR tool measures the risk of non-adherence for patients with chronic disease, as well as measuring the relative importance of thirteen behavioral drivers contributing to that risk. Over a period of four years, five different cohorts of patients in three countries and three different pathologies were studied to contribute to the elaboration and refinement of two patient-reported adherence measures: SPUR 6 and SPUR 24. This article examines the results of retrofitting of both of these tools to earlier patient cohorts as well as analyzing the pooled dataset via the use of both tools in order to further study the predictive potential of both. A further analysis was carried out using structural equation modeling both to test the structural validity of the SPUR tools and to examine both indirect and direct influence of the thirteen drivers on patient behavior.Direct comparisons of the SPUR tools to other patient-reported adherence measures across datasets and across the pooled dataset was carried out by analysis of Spearman’s ranked correlation coefficients. The structural equation modeling was carried out using path analysis based on the decision-making schema hypothesized in the foundational SPUR article.The retrofitted analysis and the pooled data analysis both support the use of SPUR 6 and SPUR 24 to assess the risk of non-adherence of patients with chronic disease with respect to other widely used patient reported adherence measures. The structural equation modeling reinforced the hypothesis that the social and psychological drivers of SPUR have a significant indirect impact on non-adherence risk via the rational and usage drivers as well as their direct impact on non-adherence risk.SPUR 6 and SPUR 24 have demonstrated predictive value in assessing the risk of patient non-adherence as compared to their predecessors as well as to other widely-used patient adherence measures, across countries and pathologies. The social and psychological drivers of SPUR seem to drive behavior largely through their influence on rational and usage factors, indicating a cognitive rationalization process . These insights have direct implications for communication strategy towards patients in efforts to enhance medication adherence.

PMID:41006489 | DOI:10.1038/s41598-025-17866-6

Sci Rep. 2025 Sep 26;15(1):33088. doi: 10.1038/s41598-025-17657-z.

ABSTRACT

This study investigates the relationship between dietary patterns and psoriasis severity using advanced machine learning (ML) techniques. The dataset, comprising 37 features including demographic, clinical and dietary features from 142 Thai psoriasis patients, exhibits moderately high dimensionality typical of clinical studies. To address limitations posed by the small sample size, a hybrid resampling strategy integrating bootstrapping with K-fold Cross-Validation (CV) was implemented. Using Random Forest (RF) and eXtreme Gradient Boosting (XGB), a total of 60 classification models were evaluated by varying train/test splits and applying multiple feature selection methods, including Least Absolute Shrinkage and Selection Operator (LASSO), Mean Decrease Accuracy (MDA), and Mean Decrease Impurity (MDI). Although bootstrapping alone sometimes resulted in overfitting, its combination with K-fold CV improved generalizability. In optimal configurations, both RF and XGB achieved sensitivity, specificity, and F1-scores exceeding 90%, alongside area under the curve (AUC) values above 95%. SHapley Additive exPlanations (SHAP) analysis revealed key dietary factors associated with increased psoriasis severity, including high-sodium foods, processed meats, alcohol, red meats, fermented products, and dark-colored vegetables. Clinically, prioritizing weight management is essential, as Body Mass Index (BMI) arose as the strongest feature of psoriasis severity. Dietary triggers identified in this study should inform comprehensive care plans. Popular Thai cuisines, especially Tom Yum Kung emerged as a potentially suitable option, while Som Tum, Pad Thai, Moo Kratha, and Khao Niao Mamuang were identified as potential triggers when consumed excessively. These findings highlight the importance of dietary moderation and personalized guidance, supporting health literacy, patient management, and smart healthcare innovations in Thailand.

PMID:41006473 | DOI:10.1038/s41598-025-17657-z