ObjectiveDiscriminating atypical hepatocellular carcinoma (HCC) from other malignancies in liver nodules classified as Liver Imaging Reporting and Data System category M (LR-M) remains a significant diagnostic challenge on conventional ultrasound examination. The LR-M category, originally intended to capture non-HCC malignancies, paradoxically contains up to 63% of atypical HCCs that deviate from classic enhancement patterns, leading to potential misdiagnosis and suboptimal treatment planning. While deep learning has shown promise in HCC diagnosis, most existing models rely exclusively on single-modality ultrasound, overlooking the diagnostic benefits of integrating complementary information from multiple imaging sources. To address this gap, we propose a novel attention-weighted tri-modal ultrasound network (TUS-Net) that integrates contrast-enhanced ultrasound (CEUS), B-mode ultrasound (BUS), and time-intensity curves (TICs) to improve diagnostic accuracy for these clinically challenging lesions. MethodsOur framework incorporates a three-dimensional convolutional neural network (C3D) backbone to extract spatiotemporal features from CEUS videos, capturing dynamic vascular patterns critical for lesion characterization. To effectively fuse complementary modalities, we introduce a dual-channel feature fusion module (DCFFM) that adaptively combines features from CEUS and BUS through channel-wise attention mechanisms, allowing the model to dynamically weigh the contribution of each modality based on diagnostic relevance. Additionally, we propose a temporal intensity feature fusion module (TIFFM) that leverages quantitative hemodynamic information from TICs to guide the model’s attention toward diagnostically critical temporal phases, such as arterial wash-in and portal venous washout. The model is further enhanced by automated lesion localization using YOLOX and class activation mapping for interpretability, ensuring that predictions align with clinically meaningful imaging features. ResultsEvaluated on a tri-modal ultrasound dataset comprising 161 patients with pathologically confirmed LR-M nodules (131 atypical HCC and 30 non-HCC malignancies), our model achieved an accuracy of 86.83%, a sensitivity of 92.50%, a specificity of 75.50%, and an AUC of 89.32% in screening atypical HCC. Compared to single-modality baselines, TUS-Net demonstrated superior specificity, a clinically critical metric given the higher risk associated with misclassifying non-HCC malignancies. Ablation studies confirmed the contribution of each module, with the full model outperforming both standard C3D and 3D ResNet backbones integrated with attention mechanisms. A reader study involving junior and senior radiologists further validated the clinical utility of AI assistance, showing consistent improvements in specificity and inter-reader consistency, particularly for less experienced clinicians. ConclusionThese results surpass existing benchmark models and demonstrate the potential of our approach to enhance diagnostic precision in clinically specific cases. By intelligently fusing multi-modal ultrasound data with attention-guided mechanisms, TUS-Net offers a reliable and interpretable tool that holds promise for improving the non-invasive diagnosis of atypical HCC in challenging LR-M liver nodules.

ObjectiveDiscriminating atypical hepatocellular carcinoma (HCC) from other malignancies in liver nodules classified as Liver Imaging Reporting and Data System category M (LR-M) remains a significant diagnostic challenge on conventional ultrasound examination. The LR-M category, originally intended to capture non-HCC malignancies, paradoxically contains up to 63% of atypical HCCs that deviate from classic enhancement patterns, leading to potential misdiagnosis and suboptimal treatment planning. While deep learning has shown promise in HCC diagnosis, most existing models rely exclusively on single-modality ultrasound, overlooking the diagnostic benefits of integrating complementary information from multiple imaging sources. To address this gap, we propose a novel attention-weighted tri-modal ultrasound network (TUS-Net) that integrates contrast-enhanced ultrasound (CEUS), B-mode ultrasound (BUS), and time-intensity curves (TICs) to improve diagnostic accuracy for these clinically challenging lesions. MethodsOur framework incorporates a three-dimensional convolutional neural network (C3D) backbone to extract spatiotemporal features from CEUS videos, capturing dynamic vascular patterns critical for lesion characterization. To effectively fuse complementary modalities, we introduce a dual-channel feature fusion module (DCFFM) that adaptively combines features from CEUS and BUS through channel-wise attention mechanisms, allowing the model to dynamically weigh the contribution of each modality based on diagnostic relevance. Additionally, we propose a temporal intensity feature fusion module (TIFFM) that leverages quantitative hemodynamic information from TICs to guide the model’s attention toward diagnostically critical temporal phases, such as arterial wash-in and portal venous washout. The model is further enhanced by automated lesion localization using YOLOX and class activation mapping for interpretability, ensuring that predictions align with clinically meaningful imaging features. ResultsEvaluated on a tri-modal ultrasound dataset comprising 161 patients with pathologically confirmed LR-M nodules (131 atypical HCC and 30 non-HCC malignancies), our model achieved an accuracy of 86.83%, a sensitivity of 92.50%, a specificity of 75.50%, and an AUC of 89.32% in screening atypical HCC. Compared to single-modality baselines, TUS-Net demonstrated superior specificity, a clinically critical metric given the higher risk associated with misclassifying non-HCC malignancies. Ablation studies confirmed the contribution of each module, with the full model outperforming both standard C3D and 3D ResNet backbones integrated with attention mechanisms. A reader study involving junior and senior radiologists further validated the clinical utility of AI assistance, showing consistent improvements in specificity and inter-reader consistency, particularly for less experienced clinicians. ConclusionThese results surpass existing benchmark models and demonstrate the potential of our approach to enhance diagnostic precision in clinically specific cases. By intelligently fusing multi-modal ultrasound data with attention-guided mechanisms, TUS-Net offers a reliable and interpretable tool that holds promise for improving the non-invasive diagnosis of atypical HCC in challenging LR-M liver nodules.

Background: s/Aims: Postoperative pancreatic fistulas (POPF) remain a major cause of morbidity and mortality following pancreatoduodenectomy (PD). Pancreatogastrostomy (PG) and pancreatojejunostomy (PJ) are the two most commonly used reconstruction techniques, yet evidence favoring one over the other is inconclusive. This study evaluates postoperative outcomes following open PD at a single institution that transitioned from PG to PJ as the preferred reconstruction method. Methods: This retrospective comparative study included patients who underwent PD between April 2005 and August 2022. Of 757 patients identified, 522 met the inclusion criteria. Propensity score matching (PSM) was performed to adjust for clinically relevant covariates. Primary endpoints were clinically relevant (CR) POPF (grade B/C) and Clavien–Dindo (CD) grade ≥ 3 POPFs. Secondary outcomes included post-pancreatectomy hemorrhage (PPH), delayed gastric emptying (DGE), systemic complications, length of hospital stay, and mortality. Results: Overall, CR-POPF and CD grade ≥ 3 POPFs occurred in 21.3% and 8.0% of patients, respectively. Thirty-day and in-hospital mortality rates were 3.1% and 4.2%. After PSM, 368 patients (184 PG and 184 PJ) were analyzed. Grade B POPFs were more frequent following PJ than PG (24.5% vs. 15.8%, p < 0.001). Although CR-POPF and CD grade ≥ 3 POPFs were numerically higher in the PJ group, differences were not statistically significant. In contrast, DGE, PPH, and in-hospital mortality were significantly higher following PG (37.0% vs. 25.0%, p = 0.025; 16.3% vs. 8.7%, p = 0.025; and 7.6% vs. 2.7%, p = 0.049, respectively). Conclusions PG was associated with a lower incidence of grade B POPFs but higher rates of DGE, PPH, and in-hospital mortality.

Introduction: Prediabetes or intermediate hyperglycemia (IH) represents a critical phase in the trajectory toward type 2 diabetes mellitus (T2DM). Recent evidence from the International Diabetes Federation (IDF) suggests that 1-hour OGTT (1HOGTT) ≥8.6 mmol/L is a more sensitive biomarker for early beta-cell dysfunction, with enhanced detection of IH and future T2DM risk. Conventional screening using hemoglobin A1c (HbA1c) or 2-Hour OGTT (2HOGTT) may delay identification of prediabetes, narrowing the window for early intervention. This study evaluates 1HOGTT as a screening tool for prediabetes among Malaysians. Methodology: This cross-sectional study enrolled adults without prior T2DM or prediabetes. Participants underwent 1HOGTT, 2HOGTT, and HbA1c, classified as per the Malaysian Clinical Practice Guideline for T2DM (6th Edition) and IDF 2024 criteria. Agreement between tests was assessed using McNemar’s test and Cohen’s kappa. Diagnostic accuracy was evaluated via receiver operating characteristic (ROC) curve analysis. Cost-effectiveness was determined using reagent cost only. Results: The majority of the 310 participants were female (71.6%), Malay (87.1%), with average age of 40. The 1HOGTT identified prediabetes in 21.6% (n = 67) compared to 17.1% (n = 53) by HbA1c and 2.6% (n = 8) by 2HOGTT. McNemar’s test confirmed statistically significant discordance between 1HOGTT and 2HOGTT (chi-square = 53.397, p <0.001): 61 participants were prediabetic by 1HOGTT but normal by 2HOGTT, versus only 2 in the reverse direction. No significant discordance was found between 1HOGTT and HbA1c (chi-square = 2.817, p = 0.093). 1HOGTT demonstrated excellent discriminatory ability against 2HOGTT (AUC = 0.908; 95% CI: 0.837–0.978), significantly outperforming HbA1c (AUC = 0.664; CI: 0.462– 0.867; DeLong p = 0.012). In a population-level cost analysis, 1HOGTT achieved lowest cost per prediabetes case detected (RM 5.79) – approximately 8.3 times and 8.1 times more cost-effective than 2HOGTT (RM 48.08) and HbA1c (RM 46.78) respectively. Conclusion 1HOGTT identifies a significant proportion of individuals with prediabetes missed by 2HOGTT and HbA1c. Incorporating 1HOGTT enhances the detection of prediabetes, is cost-effective, and enables timely preventive measures in our population with high diabetes burden.
Glucose Tolerance Test ; Glycated Hemoglobin ; Prediabetic State

Introduction: Insulin therapy is essential in the management of type 2 diabetes mellitus (T2DM), but is often associated with treatment burden, hypoglycemia, and potential overtreatment. Insulin deintensification is increasingly recommended for appropriately selected patients; however, there is limited real-world evidence to guide patient selection and to identify modifiable factors that influence successful insulin deintensification. This study aimed to identify clinical predictors, including modifiable factors, associated with successful insulin deintensification in a primary care setting. Methodology: A multicentre retrospective observational study was conducted across seven government primary care clinics in the Petaling District. Adult patients with T2DM undergoing insulin deintensification were included. Successful insulin deintensification was defined as maintenance or improvement of hemoglobin A1c following insulin discontinuation, dose reduction, or reduction in injection frequency. Paired outcomes were analyzed using the Wilcoxon signed-rank test. Between-group comparisons were performed using the Mann–Whitney U test and Chi-square or Fisher’s exact test. Multivariable logistic regression was used to identify independent predictors of successful insulin deintensification. Results: A total of 261 patients were included. Glycemic control remained stable following insulin deintensification (p = 0.334). Significant reductions in body weight (−0.41 kg, p = 0.012) and total daily insulin dose (23.1% reduction, p <0.001) were observed. Univariate analysis did not demonstrate significant differences between groups. However, multivariable logistic regression identified SGLT-2 inhibitor use (aOR 3.23, 95% CI 1.28–8.18, p = 0.013) and regular SMBG (aOR 2.00, 95% CI 1.05–3.81, p = 0.035) as independent predictors of successful insulin deintensification. Conclusion Insulin deintensification can be successfully implemented without compromising glycemic control. Identified predictors, including modifiable factors such as SGLT-2 inhibitor use and SMBG, provide clinically actionable insights to guide patient selection and treatment optimization. These findings challenge the traditional reluctance toward insulin deintensification and support a more evidence-based and individualized approach in routine clinical practice.
Primary Health Care ; Insulins

Introduction: Adequate knowledge of diabetes is essential for effective self-management and glycemic control. Emerging trends in digital platforms, video-based education offer a scalable and accessible approach to patient learning. However, evidence on the effectiveness of validated YouTube-based diabetes education in the Malaysian population remains limited. We aimed to evaluate the impact of a newly developed, tri-language YouTube diabetes education programme on clinical and knowledge outcomes. Methodology: In this multicentre study, patients with diabetes mellitus were randomly assigned to intervention (YouTube-based education) and control groups (standard care). The intervention consisted of newly developed and validated short educational videos on YouTube regarding diabetes, complications and lifestyle changes, delivered weekly via WhatsApp over 9 weeks. Videos were available in English, Malay and Chinese. The control group received standard care and counseling. Demographic data, hemoglobin A1c (HbA1c), and knowledge scores were collected at baseline and 3 months post-intervention. Results: A total of 109 patients (62.4% female; mean age 53.2 years; mean diabetes duration 16.3 years) were enrolled in the study, with similar baseline characteristics between groups. The intervention group demonstrated significantly greater improvement in glycemic control compared to the control group (HbA1c: −1.22 ± 1.36% vs. −0.24 ± 1.05%; p <0.001). While knowledge scores improved in both groups, the increase was significantly greater in the intervention group (2.66 vs. 0.80; p = 0.001). Notably, improvement in knowledge score predicted HbA1c reduction (b = 0.091; 95% CI 0.01, 0.172; p = 0.028). Conclusion The tri-language YouTube-based diabetes education programme significantly improved both glycemic control and patients’ knowledge. This accessible, scalable video-based education represents an effective strategy to enhance diabetes selfmanagement, offering flexible, on-demand learning for diverse populations.
Social Media ; Diabetes Mellitus

Introduction: Current education on diabetes self-management is time-consuming as it is conducted face-to-face. This study aimed to evaluate the effectiveness of diabetes education via mobile application on hemoglobin A1c (HbA1c). Methodology: This was a double-blinded randomized controlled study conducted in the Department of Endocrinology, Hospital Pulau Pinang and Hospital Raja Permaisuri Bainun. Participants were stratified randomized according to HbA1c to intervention group (using T-Manis) and control group. The primary objective was to assess the difference in HbA1c between the groups. Secondary objectives were to assess the difference in body weight, hospitalization related to diabetes, diabetes knowledge between groups and usability and satisfaction of T-Manis. Laboratory investigations and questionnaires were conducted at baseline, 4 months and 8 months. Results: A total of 112 participants were randomized into two groups each with 56 participants. Mean age was 51 years with similar diabetes duration of 16 years in both groups. Participants comprised of 61.6% females with ethnic distribution of 49.1% Malay, 30.4% Chinese, 19.6% Indian and 0.9% Siamese. Baseline HbA1c levels were comparable between intervention and control groups (10.04% versus 9.91%; mean difference 0.13%; 95% CI −0.56 to 0.81; p = 0.718). At 4 months, HbA1c of the intervention group demonstrated a significant reduction compared to the control group (8.93% versus 9.80%; mean difference −0.86; 95% CI −1.59 to −0.14; p = 0.02). Sustained reduction of HbA1c in the intervention group was observed at 8 months (mean difference −1.01; 95% CI −1.85 to −0.16; p = 0.02). Baseline body weight was 77 kg, with no betweengroup differences in body weight, hospitalization related to diabetes and diabetes knowledge scores, although diabetes knowledge scores improved at 4 months in the intervention group. 89.3% of participants reported satisfaction with T-Manis Conclusion The T-Manis mobile application aided in significantly improving glycemic control with high user satisfaction, supporting its feasibility as an adjunct to standard diabetes care.
Diabetes Mellitus, Type 2 ; Glycemic Control ; Mobile Applications ; Pangolins

Objective Exploring genetic differences among geographic strains of Anopheles sinensis in China. Methods Based on the mitochondrial cytochrome oxidase subunit I(COⅠ )and ribosomal the second internal transcribed spacer(ITS2), genetic distance analyses were conducted on An. sinensis samples collected nationwide. Results Genetic differentiation was observed within An. sinensis geographic strains, primarily reflected in the genetic distances of COⅠ sequences, with some distances between samples exceeding those between closely related species. However, ITS2 sequences showed relatively low levels of differentiation within geographic strains. No significant separation was observed, and gene exchange exhibited a random pattern among geographic strains. Conclusions Significant differences in genetic distance were observed among An. sinensis samples, along with evidence of interspecific continuity among sibling species. For the identification of closely related species using wild-collected Anopheles mosquitoes samples, reliance solely on commonly used barcodes such as COⅠ and ITS2 may be insufficient. Additional molecular markers, including COⅡ and D3 gene fragments, should also be considered. Caution should be exercised when using COⅠ genes to identify cryptic species, especially within Anopheles mosquitoes.

Diabetic nephropathy belongs to diabetic microvascular aseptic inflammation， which is the main cause of end-stage renal disease and can cause damage to the structure of glomeruli and tubulointerstitium. Pyroptosis is a potent pro-inflammatory model of programmed cell death mediated by gasdermin （GSDM） protein family， which is characterized by nuclear shrinkage， chromatin DNA fragmentation， plasma membrane rupture， cell swelling and deformation. Related studies have found that pyroptosis is closely related to the occurrence and development of diabetic nephropathy. Inhibition of pyroptosis to a certain extent can reduce diabetic renal injury， so the pathway of pyroptosis involved in the pathogenesis of diabetic nephropathy is the current research focus. The characteristics of multi-components， multi-targets and multi-pathways of traditional Chinese medicine have significant advantages in the treatment of diabetic nephropathy. Some components of traditional Chinese medicine， single traditional Chinese medicine and compound prescription of traditional Chinese medicine can improve diabetic renal injury by inhibiting pyroptosis pathway and related targets. Based on the molecular mechanism of pyroptosis， this paper clarifies the mechanism of pyroptosis in diabetic nephropathy by expounding the relationship between pyroptosis and its pathway targets and diabetic nephropathy. This paper summarizes the research status of traditional Chinese medicine in the treatment of diabetic nephropathy by interfering with related pyroptosis pathways in recent years， so as to provide perspectives and ideas for the prevention， early diagnosis， drug development and molecular treatment of diabetic nephropathy in the future.

Purpose: The aim of this study was to investigate the feasibility of an intelligent handheld ultrasound (US) device for assisting non-expert general practitioners (GPs) in detecting carotid plaques (CPs) in community populations. Methods: This prospective parallel controlled trial recruited 111 consecutive community residents. All of them underwent examinations by non-expert GPs and specialist doctors using handheld US devices (setting A, setting B, and setting C). The results of setting C with specialist doctors were considered the gold standard. Carotid intima-media thickness (CIMT) and the features of CPs were measured and recorded. The diagnostic performance of GPs in distinguishing CPs was evaluated using a receiver operating characteristic curve. Inter-observer agreement was compared using the intragroup correlation coefficient (ICC). Questionnaires were completed to evaluate clinical benefits. Results: Among the 111 community residents, 80, 96, and 112 CPs were detected in settings A, B, and C, respectively. Setting B exhibited better diagnostic performance than setting A for detecting CPs (area under the curve, 0.856 vs. 0.749; P<0.01). Setting B had better consistency with setting C than setting A in CIMT measurement and the assessment of CPs (ICC, 0.731 to 0.923). Moreover, measurements in setting B required less time than the other two settings (44.59 seconds vs. 108.87 seconds vs. 126.13 seconds, both P<0.01). Conclusion Using an intelligent handheld US device, GPs can perform CP screening and achieve a diagnostic capability comparable to that of specialist doctors.