In view of the complex structure of diabetic retinopathy and the large differences in the scales of different lesions,a novel network which integrates deformable convolution and attention mechanism is proposed for automatic diabetic retinopathy multi-lesion segmentation.Specifically,deformable convolution Haar wavelet transform encoder takes place of the original convolutional downsampling encoder to adapt to the irregular shape changes of lesions and extract effective feature information;a dense feature perception and aggregation module is introduced at the bottleneck layer to extract multi-scale features by aggregating multiple receptive fields,thus enhancing deep semantic information;and finally,in order to fully integrate the decoder output and improve the recognition accuracy of edge information,a multi scale adaptive fusion module is used to weight the decoder output of each layer for obtaining the most accurate segmentation feature map.The validation of hard percolation,bleeding point,and soft percolation segmentations on the DDR-RLS dataset reveals that the proposed network shows increases of 0.026 2,0.051 8 and 0.046 5 in IoU coefficient,0.027 1,0.058 1 and 0.050 4 in Dice coefficient,and 0.0423,0.0691 and 0.0734 in AUPR value,as compared with the original Unet.

Dendrobium moniliforme (D. moniliforme) is a traditional medicinal herb widely cultivated in Asia. Flavonoids, one of the largest groups of secondary metabolites in plants, are significant medicinal components in Dendrobium species. Several subgroups of R2R3-MYB proteins have been validated to directly regulate flavonoid biosynthesis. Using PacBio sequencing technology, we assembled a high-quality chromosome-level D. moniliforme genome with a total length of 1.20 Gb and a contig N50 of 3.97 Mb. The BUSCO assessment of genome annotation was 91.4%. By integrating the genome and transcriptome, we identified biosynthesis pathway enzyme genes related to flavonoids, polysaccharides, carotenoids, and alkaloids. A total of 90 R2R3-MYBs were identified in D. moniliforme and classified into 21 subgroups. Studies on the functions of R2R3-MYB transcription factors revealed that R2R3-MYB in SG6 can up-regulate flavonoid biosynthesis. Various validation experiments, including subcellular localization, transient overexpression, UPLC-MS/MS, HPLC, yeast one-hybrid, and dual-luciferase assays, demonstrated that DMYB69 directly up-regulates the expression of enzyme genes involved in flavonoid biosynthesis, increasing the content of flavonoids such as anthocyanin, flavone, and flavonol. Additionally, DMYB44 was shown to directly up-regulate the expression of carotenoid biosynthesis enzyme genes, thereby increasing carotenoid content. This study provides an essential genome resource and theoretical basis for molecular breeding research in D. moniliforme.

ObjectiveTo investigate the triglyceride-glucose (TyG) index and the level of serum homocysteine (Hcy) in association with the incidence of stroke in type 2 diabetes mellitus (T2DM) patients. MethodsBased on the chronic disease risk factor surveillance cohort in Pudong New Area, Shanghai, excluding those with stroke in baseline survey, T2DM patients who joined the cohort from January 2016 to October 2020 were selected as the research subjects. During the follow-up period, a total of 318 new-onset ischemic stroke patients were selected as the case group, and a total of 318 individuals matched by gender without stroke were selected as the control group. The Cox proportional hazards regression model was used to adjust for confounding factors and explore the serum TyG index and the Hcy biochemical indicator in association with the risk of stroke. ResultsThe Cox proportional hazards regression results showed that after adjusting for confounding factors, the risk of stroke in T2DM patients with 10 μmol·L⁻¹-¹ and Hcy>15 μmol·L⁻¹ was 1.532 times (95%CI: 1.017‒2.309 times, P=0.041) and 1.738 times (95%CI: 1.119‒2.699 times, P=0.014) higher respectively, compared to T2DM patients with Hcy≤10 μmol·L⁻¹. T2DM patients with TyG>9.074 had 1.396 times higher risk of stroke (95%CI: 1.091‒1.787, P=0.008) compared to those with TyG≤9.074. The study found that urea and α1-antitrypsin (α1-AT) were independent risk factors for the incidence of stroke in T2DM patients. For every unit increase in urea andα1-AT, the risk of stroke in T2DM patients increased by 233.6% (HR=3.336, 95%CI: 1.588‒7.009, P=0.001) and 11.3% (HR=1.113, 95%CI: 1.028‒1.205, P=0.008), respectively. Cumulative risk curves showed that Hcy>10 μmol·L⁻¹ and TyG>9.074 accelerated the time to stroke occurrence in T2DM patients. ConclusionElevated levels of Hcy>10 μmol·L⁻¹ and a higher TyG index are risk factors for stroke in T2DM patients. Effective interventions for Hcy and TyG should be conducted for diabetic patients to reduce the incidence of stroke.

OBJECTIVE To comprehensively evaluate the efficacy， safety and cost-effectiveness of insulin icodec in treating type 2 diabetes mellitus （T2DM）， providing evidence-based guidance for new drug selection in hospital and clinical medication decision-making. METHODS PubMed， Cochrane Library， Embase， CNKI， Wanfang， VIP and foreign health technology assessment （HTA） websites were searched by using rapid health technology assessment from inception to 15 July 2025 for systematic reviews/meta-analyses， pharmacoeconomic studies， and HTA reports on insulin icodec in the treatment of T2DM. After data extraction and quality assessment， the findings of the included studies were analyzed descriptively. RESULTS Ten systematic reviews/meta-analyses and three pharmacoeconomic studies were included. Among them， 4 systematic reviews/meta-analyses were of high quality； the overall quality of the 3 pharmacoeconomic studies was relatively good. Regarding efficacy， insulin icodec was superior to once-daily basal insulin in reducing glycated hemoglobin （HbA1c） and in achieving the target of HbA1c＜7% （P＜0.05）. No significant differences were observed between icodec insulin and comparators in lowering fasting plasma glucose （P＞0.05）. For safety， insulin icodec did not increase the incidence of any adverse events （AEs）， serious AEs， clinically significant hypoglycemia （random glucose＜3 mmol/L）， injection-site reactions， or allergic reactions， compared with once-daily basal insulin overall （P＞ 0.05）； however， insulin icodec was associated with a significant increase in body weight （P＜0.05）. Domestic economic evaluations indicated that insulin icodec was more cost-effective than insulin glargine and insulin degludec when its annual costs were in the range of 784.90-1 145.96 and 597.66-736.34 US dollars， respectively. CONCLUSIONS Insulin icodec demonstrates favorable efficacy and safety profiles in the treatment of T2DM； however， attention should be paid to the risk of weight gain. Under China’s healthcare system， insulin icodec demonstrates greater economic value only when the patient’s weekly required basal insulin dose falls within a specific range，and clinical practice requires individualization.

The research on medical image segmentation is of great significance in advancing efficient and accurate automated image processing techniques.To address the problem of inaccurate segmentation results caused by significant variations in organ tissue shapes and blurred boundaries present in medical images,a novel network named MFMANet is proposed.Built upon a"U"-shaped architecture,the network integrates multi-scale information fusion modules and multi-attention modules.Specifically,multi-scale information modules capture multi-scale information in the shallow layers of the network to bridge the semantic gap between encoder and decoder features,thereby enhancing the network's ability to handle large variations in organ sizes.Regarding the issue of blurred boundaries,multi-attention mechanism utilizes Swin Transformer as the deep encoder-decoder network,employing channel and spatial attention instead of traditional skip connections to achieve finer feature extraction.Experimental results on the ACDC and Synapse public datasets show that the proposed method achieves improvements of 1.51%and 1.29%in Dice similarity coefficient as compared with MTUNet,fully demonstrating its effectiveness in enhancing segmentation network accuracy.

The diagnosis of Meige's syndrome predominantly relies on the clinical assessment by physicians.Given the complexity and similarity of its symptoms to other neurological disorders,the diagnosis is crucial for both doctors and patients.Herein a detection dataset for Meige's syndrome is compiled from video recordings of 31 patients,and an automated diagnostic system for Meige's syndrome(MS-Net)applicable to untrimmed videos is developed.The system utilizes RetinaNet and UNet3+to construct temporal detection and segmentation branches for multi-scale feature extraction and temporal segmentation,obtains probability vectors for detection windows and the probability of disease onset per frame via the decoding of temporal detection and segmentation branches,and finally generates a refined probability for each window by processing the probability predictions from both branches using a multi-layer perceptron.The model performance is optimized using additional loss functions and data augmentation techniques,operating on features interpretable by clinical physicians.MS-Net can assist in the diagnosis of Meige's syndrome,improving the accuracy,convenience,and efficiency of the early diagnosis.The comparison of MS-Net with other state-of-the-art networks indicates that MS-Net achieves comparable performance in terms of average precision while utilizing interpretable features required in clinical practice.

The diagnosis of Meige's syndrome predominantly relies on the clinical assessment by physicians.Given the complexity and similarity of its symptoms to other neurological disorders,the diagnosis is crucial for both doctors and patients.Herein a detection dataset for Meige's syndrome is compiled from video recordings of 31 patients,and an automated diagnostic system for Meige's syndrome(MS-Net)applicable to untrimmed videos is developed.The system utilizes RetinaNet and UNet3+to construct temporal detection and segmentation branches for multi-scale feature extraction and temporal segmentation,obtains probability vectors for detection windows and the probability of disease onset per frame via the decoding of temporal detection and segmentation branches,and finally generates a refined probability for each window by processing the probability predictions from both branches using a multi-layer perceptron.The model performance is optimized using additional loss functions and data augmentation techniques,operating on features interpretable by clinical physicians.MS-Net can assist in the diagnosis of Meige's syndrome,improving the accuracy,convenience,and efficiency of the early diagnosis.The comparison of MS-Net with other state-of-the-art networks indicates that MS-Net achieves comparable performance in terms of average precision while utilizing interpretable features required in clinical practice.

In view of the complex structure of diabetic retinopathy and the large differences in the scales of different lesions,a novel network which integrates deformable convolution and attention mechanism is proposed for automatic diabetic retinopathy multi-lesion segmentation.Specifically,deformable convolution Haar wavelet transform encoder takes place of the original convolutional downsampling encoder to adapt to the irregular shape changes of lesions and extract effective feature information;a dense feature perception and aggregation module is introduced at the bottleneck layer to extract multi-scale features by aggregating multiple receptive fields,thus enhancing deep semantic information;and finally,in order to fully integrate the decoder output and improve the recognition accuracy of edge information,a multi scale adaptive fusion module is used to weight the decoder output of each layer for obtaining the most accurate segmentation feature map.The validation of hard percolation,bleeding point,and soft percolation segmentations on the DDR-RLS dataset reveals that the proposed network shows increases of 0.026 2,0.051 8 and 0.046 5 in IoU coefficient,0.027 1,0.058 1 and 0.050 4 in Dice coefficient,and 0.0423,0.0691 and 0.0734 in AUPR value,as compared with the original Unet.

The research on medical image segmentation is of great significance in advancing efficient and accurate automated image processing techniques.To address the problem of inaccurate segmentation results caused by significant variations in organ tissue shapes and blurred boundaries present in medical images,a novel network named MFMANet is proposed.Built upon a"U"-shaped architecture,the network integrates multi-scale information fusion modules and multi-attention modules.Specifically,multi-scale information modules capture multi-scale information in the shallow layers of the network to bridge the semantic gap between encoder and decoder features,thereby enhancing the network's ability to handle large variations in organ sizes.Regarding the issue of blurred boundaries,multi-attention mechanism utilizes Swin Transformer as the deep encoder-decoder network,employing channel and spatial attention instead of traditional skip connections to achieve finer feature extraction.Experimental results on the ACDC and Synapse public datasets show that the proposed method achieves improvements of 1.51%and 1.29%in Dice similarity coefficient as compared with MTUNet,fully demonstrating its effectiveness in enhancing segmentation network accuracy.