Aging has emerged as a cutting edge and hotspot in global life science field， with anti-aging and geriatric disease prevention and treatment becoming critical issues urgently demanding solutions in international medical communities. In the face of the challenge of accelerating global population aging， in-depth exploration of aging mechanisms and the development of effective intervention strategies hold significant scientific and clinical value. This study supported by the national key research and development program of China， employed the essence-Qi-spirit theory of Qiluo doctrine as its guiding framework， focusing on the key scientific issue of the core traditional Chinese pathogenesis of aging， namely "depletion of kidney essence， deficiency of primordial Qi， and impairment of body and spirit". The treatment principle of "tonifying the kidney to replenish essence， harmonizing Yin and Yang， warming and invigorating primordial Qi， and nourishing the body and spirit" was established. Centered on holistic aging， systemic aging， and aging-related diseases， the research integrated multidisciplinary research approaches to construct multi-modal aging models and a multi-dimensional evaluation system， and it utilized multi-omics technologies to deeply analyze aging mechanisms. By systematically reviewing historical kidney-tonifying and anti-aging formulas and combining big data with artificial intelligence technologies， an information database of anti-aging traditional Chinese medicine substance was developed to reveal the differences and synergistic effects of various treatment methods and formulas on anti-aging. Based on this treatment method， the research integrated two millennia of kidney-tonifying medicinal experience to develop the innovative anti-aging traditional Chinese medicine， namely Bazhi Bushen capsules. It was validated that this capsule can delay holistic and systemic aging through multiple targets and mechanisms， thereby elucidating the scientific connotation of the essence-Qi-spirit theory of Qiluo doctrine in guiding anti-aging research from multiple dimensions and providing robust support for leveraging the advantages of traditional Chinese medicine to occupy the commanding heights of international anti-aging research.

BACKGROUND:Rapid and accurate segmentation of brain tissue in medical images is of great significance for three-dimensional biomechanical modeling and diagnosis of craniocerebral injuries.Currently,artificial intelligence(AI)-based baseline models exhibit excellent generalization capabilities on large-scale datasets.However,due to the specificity and complexity of craniocerebral tissues,these models have certain limitations in their application to craniocerebral tissue segmentation.Additionally,the scarcity of craniocerebral tissue samples makes it difficult for baseline models to achieve precise segmentation results through fine-tuning.OBJECTIVE:To construct a craniocerebral tissue segmentation model based on texture feature retrieval enhancement to improve segmentation accuracy under a small number of samples.METHODS:Segment Anything in Medical Images(MedSAM)model was selected as the basic framework,and texture features were combined with deep learning to build a brain tissue segmentation model based on texture feature retrieval enhancement(DP-MedSAM).Dice Coefficient and mean intersection over union(MIoU)were selected to evaluate the efficiency of image segmentation results.In comparison with the original MedSAM model,the ablation experiment systematically evaluated the influence of key components on the model performance.The sensitivities of MedSAM,the Segment Anything Model(SAM)for medical image segmentation(SAM-Med2D)and DP-MedSAM in the mandible,left optic nerve,and left parotid gland were compared.RESULTS AND CONCLUSION:(1)By verifying the impact of the number of point prompts on segmentation results on the HaN-Seg dataset,the experimental results indicated that the optimal Dice score was achieved with the addition of three points.(2)DP-MedSAM demonstrated performance improvements compared with MedSAM and SAM-Med2D on two datasets(HaN and Public Domain Database for Computational Anatomy).Especially on the Public Domain Database for Computational Anatomy dataset,in terms of the MIoU metric,DP-MedSAM outperformed MedSAM by 6.59％and SAM-Med2D by 37.35％;in terms of the Dice metric,DP-MedSAM outperformed MedSAM and SAM-Med2D by 4.34％and 25.32％,respectively.(3)The ablation experiment results showed that removing the texture feature extraction module in the DP-MedSAM model,relying solely on original image features,led to a significant decrease in results on the test set.Furthermore,removing the vector cache database and its retrieval enhancement function from the model,which deprived the ability of the model to perform similarity retrieval using an external knowledge base,further reduced model performance.(4)Under conditions of limited data resources,the DP-MedSAM model outperformed the other two models in all evaluation metrics.The DP-MedSAM model performed excellently when processing simple and moderately difficult samples,demonstrating a clear advantage over the other two models and indicating good generalization ability.Processing the fine structures of difficult samples placed higher demands on the model's segmentation capabilities.Although the performance of the DP-MedSAM model declined slightly,it still outperformed the other two models.(5)This study proposes an innovative craniocerebral tissue segmentation model,DP-MedSAM,which improves the baseline model's performance in capturing local details and global structural information in medical images by introducing target region texture feature extraction.Through vector similarity retrieval technology,DP-MedSAM can retrieve the feature vector most similar to the current target region from a pre-constructed vector database,providing more precise guiding information for the segmentation process.

Inteins are unique protein insertion sequences capable of self-excision, enabling the covalent ligation of flanking extein peptides via amide bond formation. This process proceeds spontaneously without requiring external enzymes, cofactors, or chemical reagents, granting inteins exceptional biocompatibility and traceless performance in protein engineering applications. Split inteins represent a specialized and versatile subclass whose splicing domains are encoded by two separate gene fragments rather than a single continuous open reading frame. These fragments, known as the N-terminal (IntN) and C-terminal (IntC) split inteins, associate through non-covalent interactions including hydrophobic forces, hydrogen bonds, and van der Waals forces to assemble into an active three-dimensional structure, which then drives efficient extein ligation and enables protein trans-splicing. Protein trans-splicing mediated by split inteins has become a cornerstone for traceless protein ligation owing to its high specificity and irreversibility, fundamentally reshaping strategies for protein modification, assembly, and functional regulation. Compared with traditional chemical ligation methods, split intein systems require no complex chemical derivatization of peptide fragments and can operate efficiently at micromolar concentrations under physiological conditions, thus avoiding structural and functional damage caused by organic reagents. In contrast to enzymatic ligation tools such as sortase, split inteins eliminate the need for additional enzymes or cofactors, simplifying reaction systems, reducing costs, and minimizing non-specific side products. These distinctive advantages render split inteins highly promising for applications in chemical biology, synthetic biology, and biopharmaceutical development. In recent years, deepened mechanistic understanding has established structure-guided rational design as the primary approach to overcoming key limitations of split inteins, including intrinsic aggregation propensity, strict extein sequence dependence, and limited splicing efficiency. Bioinformatic tools have been used to identify aggregation-prone regions in the IntN fragment, and site-directed mutagenesis of hydrophobic residues, relocation of split sites, or removal of misfolding-prone sequences has substantially reduced in vitro aggregation and improved soluble expression and assembly activity. Rational engineering of catalytic residues and adjacent flexible loops has relaxed strict amino acid preferences at extein junctions, enhancing sequence tolerance and reducing the risk of functional impairment in target proteins. Consensus design based on multiple sequence alignments has yielded ultra-fast splicing variants such as Cfa DnaE and Cat-TerL, which exhibit significantly accelerated kinetics and improved tolerance to denaturing conditions. Meanwhile, advances in structural biology have further clarified the conformational dynamics and catalytic mechanisms of splicing, supporting the precise design of high-performance intein modules. On this basis, electrostatic interaction tuning and metagenomic screening have yielded multiple mutually orthogonal split intein pairs, enabling selective multi-fragment protein ligation and providing new routes for the efficient synthesis of large multi-domain functional proteins. With these engineered split inteins offering continuously improved performance and expanded applicability, protein trans-splicing has been widely applied in numerous cutting-edge areas of protein research and biomedicine. In gene delivery, split intein-based systems overcome the packaging limit of adeno-associated viral vectors, enabling the accurate reconstitution of large therapeutic proteins and base editors in target cells, thereby enhancing the efficacy and scope of gene therapy for genetic diseases. In internal protein sequence editing, split inteins mediate precise sequence replacement and modification in flexible regions or loops of target proteins, without the need for complex multi-step ligation and protein refolding involved in traditional protein semisynthesis. In protein-protein interaction studies, intein-mediated splicing covalently captures transient and weak intracellular complexes, enabling sensitive, high-throughput interaction detection and drug screening. In synthetic biology, conditionally controllable splicing systems support the construction of diverse intracellular and cell-surface biological logic gates for the precise regulation of cellular behavior. In mechanistic biochemical research, split inteins enable photocatalytic proximity labeling and site-specific tagging, allowing the preparation of homogeneous protein samples carrying precise post-translational modifications such as ubiquitination and polyglutamylation for chromatin interactome analysis and epigenetic studies. Moreover, covalent trapping strategies using split inteins stabilize transient enzymatic intermediates, providing unprecedented insights into molecular mechanisms such as nucleosome ubiquitination that are difficult to elucidate using conventional methods. This review systematically summarizes key technological advances in split inteins over the past decade, highlighting engineering strategies, mechanistic insights, and the development of orthogonal components. It comprehensively surveys emerging applications at the frontiers of protein research, analyzes current core challenges, and proposes future directions, particularly emphasizing artificial intelligence-driven de novo design and novel splicing pathways to break existing technical bottlenecks. By enabling traceless, efficient, and versatile protein manipulation, split inteins continue to serve as indispensable tools that drive innovation in protein engineering and fundamental life science research.

Acute kidney injury (AKI) is a prevalent and life-threatening clinical syndrome characterised by a rapid decline in renal function and diverse pathological etiologies. The condition has been demonstrated to be associated with elevated mortality rates and an increased risk of progression to chronic kidney disease. At present, clinicians depend heavily on conventional functional markers, such as serum creatinine and urine output, for the diagnosis and staging of the disease. It is evident that these conventional indicators characteristically manifest a considerable temporal delay and only undergo modification subsequent to considerable tissue damage. This severely restricts the timeframe for early detection and timely therapeutic intervention. Furthermore, standard markers fail to provide specific biological information regarding the underlying cellular injury mechanisms. The utilisation of advanced probe technologies in molecular imaging offers a robust alternative to overcome these inherent diagnostic limitations.This comprehensive review systematically evaluates recent progress in the design and application of two primary categories of molecular imaging tools for acute kidney disease, specifically reactive probes and enzyme-activated probes. Reactive probes are engineered to specifically interact with redox-active chemical species, including hydrogen peroxide, peroxynitrite, hypochlorous acid, and sulfur dioxide. Because oxidative stress constitutes a primary early event in acute renal tubular damage, these probes enable researchers and clinicians to visualize early cellular injury and radical accumulation well before global renal functional decline becomes evident. We discuss the application of these reactive probes across multiple imaging modalities including fluorescence imaging, magnetic resonance imaging (MRI), positron emission tomography (PET), and photoacoustic techniques. Photoacoustic imaging combines high spatial resolution with deep tissue penetration and has successfully demonstrated the ability to provide diagnostic alerts up to 12 h before any detectable rise in serum creatinine levels. Additionally, specific reactive probes have shown promising translational potential when tested by high-throughput screening in clinical human urine samples. Enzyme-activated probes target the specific catalytic activity of disease-relevant enzymes. These include well-documented renal tubular structural biomarkers such as NAG, GGT, and ALP, along with apoptosis-related caspases and specific nitroreductases. By responding only to enzymatic cleavage, these tools provide highly specific and pathology-directed imaging readouts. Recent structural design strategies in this field have advanced significantly beyond single-enzyme detection. Researchers are now focusing on sophisticated dual-target recognition to minimize background noise, multimodal integration to cross-validate imaging signals, and theranostic applications where probes simultaneously deliver diagnostic feedback and therapeutic agents to injured tissues. Nanotechnology serves as a fundamental enabler for realizing these advanced probe functions. By precisely optimizing nanoparticle parameters such as hydrodynamic size, surface charge, and targeting ligands, researchers can achieve amplified signal output, highly precise kidney delivery, and protection against premature degradation in the systemic circulation. For example, modifying surface charges can significantly enhance the active uptake of nanoprobes by damaged renal tubular epithelial cells.While preclinical probe development has progressed rapidly, moving these technologies into routine clinical practice remains a major challenge. We analyze the translational feasibility and current obstacles from biological, technological, and regulatory perspectives. Although biological targets such as KIM-1, FAP, and ALP have been validated in extensive patient cohorts, practical barriers severely limit their immediate clinical application. These obstacles involve complex changes in in vivo pharmacokinetics. During an acute injury episode, the extreme drop in the glomerular filtration rate alters probe clearance and can cause unwanted systemic accumulation or confusing background imaging signals. Other major hurdles include a lack of comprehensive long-term toxicity data and the absence of standardized manufacturing protocols to ensure batch-to-batch consistency. Future successful translation will require rigorous multi-center clinical studies to confirm the true diagnostic value of these probes over traditional markers. Researchers must also establish strict standardization of imaging procedures and comprehensive safety evaluations. Ultimately, this review provides a thorough reference framework for designing clinically translatable molecular probes and building a precision diagnostic imaging system for acute kidney injury.

Glycogen synthase 1 (GYS1) is the rate-limiting enzyme responsible for glycogen synthesis in skeletal muscle, heart, brain, and other extrahepatic tissues, playing a central role in systemic energy homeostasis. The human GYS1 gene maps to chromosome 19q13.33, comprises 16 exons, and encodes a 737-amino-acid polypeptide that is highly conserved across mammals. GYS1 activity is subject to multilayered and precisely coordinated regulation. At the transcriptional level, the GYS1 promoter contains a hypoxia response element (HRE) that mediates HIF-1α-dependent induction under low-oxygen conditions, as well as a muscle-specific enhancer harboring MEF2 and MyoD binding sites that confers tissue-restricted expression. At the post-translational level, a hierarchical phosphorylation cascade serves as the primary activity switch: glycogen synthase kinase 3β (GSK3β) sequentially phosphorylates four C-terminal serine residues following casein kinase II priming, while protein kinase A (PKA) and AMP-activated protein kinase (AMPK) provide parallel inhibitory inputs at both N- and C-terminal sites. Dephosphorylation and reactivation are mediated by protein phosphatase 1 (PP1) through tissue-specific glycogen-targeting regulatory subunits such as PPP1R3A and PPP1R3B, which anchor PP1 to glycogen particles and direct its activity toward GYS1. The allosteric activator glucose-6-phosphate (G6P) binds at the dimer interface, simultaneously enhancing catalytic efficiency and promoting dephosphorylation susceptibility, thereby establishing a feed-forward activation loop that couples substrate availability to glycogen synthesis. Beyond phosphorylation, GYS1 is regulated by ubiquitination (mediated by the E3 ligase PJA1), acetylation, O-linked β-N-acetylglucosamine (O-GlcNAc) modification, and SUMOylation, which collectively modulate protein stability, subcellular localization, and protein-protein interactions. Epigenetic mechanisms, including CpG island methylation and histone acetylation dynamics, govern chromatin accessibility at the GYS1 locus, while muscle-specific microRNAs such as miR-1 and miR-206 fine-tune GYS1 expression at the post-transcriptional level. Dysregulation of GYS1 has been identified as a central pathogenic driver in a spectrum of human diseases. In inherited glycogen storage disorders—including Lafora disease, adult polyglucosan body disease (APBD), and Pompe disease—loss of upstream regulatory control leads to GYS1 hyperactivation and the accumulation of structurally abnormal or excessive glycogen, resulting in progressive neurodegeneration, myopathy, and multiorgan dysfunction. In type 2 diabetes mellitus (T2DM), impaired insulin signaling through the PI3K-AKT-GSK3β axis maintains GYS1 in a hyperphosphorylated inactive state in skeletal muscle, compromising postprandial glucose disposal and exacerbating hyperglycemia. In oncology, GYS1 exhibits context-dependent roles across multiple cancer types. In hepatocellular carcinoma, FMO2⁺ cancer-associated fibroblasts stabilize GYS1 by competitively inhibiting PJA1-mediated ubiquitination, and stabilized GYS1 subsequently activates NF‑κB/CCL19 signaling to promote tertiary lymphoid structure formation and enhance anti-PD-1 immunotherapy responsiveness. In clear cell renal cell carcinoma, GYS1 promotes tumor progression through non-canonical NF‑κB pathway activation via the scaffold protein RPS27A. In triple-negative breast cancer, GYS1 has been identified as a trigger of disulfidptosis and an activator of NF-κB signaling through non-enzymatic facilitation of IκBα degradation. In colorectal cancer, mitochondrial fission deficiency drives AMPK-dependent GYS1 upregulation and glycogen accumulation as a compensatory survival mechanism, while in cervical cancer, GYS1-maintained glycogen reserves fuel the pentose phosphate pathway to generate NADPH for ROS clearance, thereby conferring cisplatin resistance in cancer stem cells. Therapeutic strategies targeting GYS1 have gained substantial momentum across these disease contexts. For glycogen storage disorders, antisense oligonucleotides, small interfering RNAs (e.g., ABX1100), and small-molecule inhibitors (e.g., MZ-101) have demonstrated preclinical and early clinical efficacy in reducing pathological glycogen accumulation. For T2DM, pharmacological activation of GYS1 through GSK3β inhibition or enhancement of PP1-mediated dephosphorylation is being explored to restore insulin-stimulated glycogen synthesis. In cancer, GYS1-directed interventions—including targeted silencing to sensitize tumors to chemotherapy and immune microenvironment modulation to enhance immunotherapy—represent emerging precision oncology approaches. This review provides a comprehensive and integrated account of GYS1 gene structure, tissue-specific distribution, regulatory networks, and pathogenic roles in metabolic disorders and malignancies, with the aim of establishing a theoretical framework for the development of GYS1-targeted precision therapies.

Background Under intensive dairy farming conditions, Enterococcus spp. can be transmitted between animals, farm workers, and the environment via multiple vectors such as feces, soil, water, air, and farming equipment, posing a potential threat to public health. Objective To elucidate the prevalence, distribution, and antimicrobial resistance profiles of Enterococcus faecalis (E. faecalis) and Enterococcus faecium (E. faecium) among farm workers, dairy cattle, and the farm environment in Xinjiang, and to assess the risk of their cross-host transmission. Methods From May 2024 to January 2025, a total of 317 samples were collected from 11 large-scale dairy farms in Xinjiang, China, including feces from farm workers (n=130) and dairy cattle (n=154), and environmental samples (n=33). E. faecalis and E. faecium were isolated and identified, followed by antimicrobial susceptibility testing and multilocus sequence typing (MLST) to analyze their molecular characteristics. Results A total of 183 Enterococcus isolates were obtained (66 E. faecalis and 117 E. faecium isolated). The isolation rates of both species showed statistically significant differences among the three sources (χ²=29.21, P=0.003). Antimicrobial resistance analysis revealed that E. faecalis generally exhibited higher resistance rates across multiple antibiotic classes than E. faecium. High resistance to rifampicin was observed across all sources (50.00%–81.25%), with statistical variation among origins (χ²=8.03, P=0.024). Multidrug-resistant strains accounted for 69.10% of the isolates. Multidrug resistance patterns in E. faecium varied significantly by source (χ²=27.19, P=0.014), and one isolate displayed resistance to eight antibiotic classes. MLST indicated high genetic diversity; E. faecalis was dominated by ST472 and ST227 of which the distrubution was significantly different among sources, while E. faecium primarily clustered into clonal complexes CC94 (centered on ST94) and CC17 (centered on ST22). Conclusion Resistant Enterococcus strains exhibit cross-transmission among farm workers, animals, and the environment. Under the "One Health" framework, standardized farming protocols and prudent antimicrobial use are essential to disrupt the transmission chain of resistant clones and mitigate the spread of antimicrobial resistance at its source.

Background Under intensive dairy farming conditions, Enterococcus spp. can be transmitted between animals, farm workers, and the environment via multiple vectors such as feces, soil, water, air, and farming equipment, posing a potential threat to public health. Objective To elucidate the prevalence, distribution, and antimicrobial resistance profiles of Enterococcus faecalis (E. faecalis) and Enterococcus faecium (E. faecium) among farm workers, dairy cattle, and the farm environment in Xinjiang, and to assess the risk of their cross-host transmission. Methods From May 2024 to January 2025, a total of 317 samples were collected from 11 large-scale dairy farms in Xinjiang, China, including feces from farm workers (n=130) and dairy cattle (n=154), and environmental samples (n=33). E. faecalis and E. faecium were isolated and identified, followed by antimicrobial susceptibility testing and multilocus sequence typing (MLST) to analyze their molecular characteristics. Results A total of 183 Enterococcus isolates were obtained (66 E. faecalis and 117 E. faecium isolated). The isolation rates of both species showed statistically significant differences among the three sources (χ²=29.21, P=0.003). Antimicrobial resistance analysis revealed that E. faecalis generally exhibited higher resistance rates across multiple antibiotic classes than E. faecium. High resistance to rifampicin was observed across all sources (50.00%–81.25%), with statistical variation among origins (χ²=8.03, P=0.024). Multidrug-resistant strains accounted for 69.10% of the isolates. Multidrug resistance patterns in E. faecium varied significantly by source (χ²=27.19, P=0.014), and one isolate displayed resistance to eight antibiotic classes. MLST indicated high genetic diversity; E. faecalis was dominated by ST472 and ST227 of which the distrubution was significantly different among sources, while E. faecium primarily clustered into clonal complexes CC94 (centered on ST94) and CC17 (centered on ST22). Conclusion Resistant Enterococcus strains exhibit cross-transmission among farm workers, animals, and the environment. Under the "One Health" framework, standardized farming protocols and prudent antimicrobial use are essential to disrupt the transmission chain of resistant clones and mitigate the spread of antimicrobial resistance at its source.

BACKGROUND:Rapid and accurate segmentation of brain tissue in medical images is of great significance for three-dimensional biomechanical modeling and diagnosis of craniocerebral injuries.Currently,artificial intelligence(AI)-based baseline models exhibit excellent generalization capabilities on large-scale datasets.However,due to the specificity and complexity of craniocerebral tissues,these models have certain limitations in their application to craniocerebral tissue segmentation.Additionally,the scarcity of craniocerebral tissue samples makes it difficult for baseline models to achieve precise segmentation results through fine-tuning.OBJECTIVE:To construct a craniocerebral tissue segmentation model based on texture feature retrieval enhancement to improve segmentation accuracy under a small number of samples.METHODS:Segment Anything in Medical Images(MedSAM)model was selected as the basic framework,and texture features were combined with deep learning to build a brain tissue segmentation model based on texture feature retrieval enhancement(DP-MedSAM).Dice Coefficient and mean intersection over union(MIoU)were selected to evaluate the efficiency of image segmentation results.In comparison with the original MedSAM model,the ablation experiment systematically evaluated the influence of key components on the model performance.The sensitivities of MedSAM,the Segment Anything Model(SAM)for medical image segmentation(SAM-Med2D)and DP-MedSAM in the mandible,left optic nerve,and left parotid gland were compared.RESULTS AND CONCLUSION:(1)By verifying the impact of the number of point prompts on segmentation results on the HaN-Seg dataset,the experimental results indicated that the optimal Dice score was achieved with the addition of three points.(2)DP-MedSAM demonstrated performance improvements compared with MedSAM and SAM-Med2D on two datasets(HaN and Public Domain Database for Computational Anatomy).Especially on the Public Domain Database for Computational Anatomy dataset,in terms of the MIoU metric,DP-MedSAM outperformed MedSAM by 6.59％and SAM-Med2D by 37.35％;in terms of the Dice metric,DP-MedSAM outperformed MedSAM and SAM-Med2D by 4.34％and 25.32％,respectively.(3)The ablation experiment results showed that removing the texture feature extraction module in the DP-MedSAM model,relying solely on original image features,led to a significant decrease in results on the test set.Furthermore,removing the vector cache database and its retrieval enhancement function from the model,which deprived the ability of the model to perform similarity retrieval using an external knowledge base,further reduced model performance.(4)Under conditions of limited data resources,the DP-MedSAM model outperformed the other two models in all evaluation metrics.The DP-MedSAM model performed excellently when processing simple and moderately difficult samples,demonstrating a clear advantage over the other two models and indicating good generalization ability.Processing the fine structures of difficult samples placed higher demands on the model's segmentation capabilities.Although the performance of the DP-MedSAM model declined slightly,it still outperformed the other two models.(5)This study proposes an innovative craniocerebral tissue segmentation model,DP-MedSAM,which improves the baseline model's performance in capturing local details and global structural information in medical images by introducing target region texture feature extraction.Through vector similarity retrieval technology,DP-MedSAM can retrieve the feature vector most similar to the current target region from a pre-constructed vector database,providing more precise guiding information for the segmentation process.

Abstract To clarify the service targets, service contents, and safeguard mechanisms of the basic public service program for student vision health management, the study grounded in the current status of child and adolescent myopia prevention and control in China and integrated with the practical experience and innovative practices of Wuhan National Demonstration Area for Student Vision Health Management, takes education as the main thread and schools as the base, consolidating resources from multiple stakeholders including education administrative departments, schools, families, and professional technical service institutions. Centering on four core components—vision health education, monitoring and record keeping, risk prevention and control, and dynamic management, the study constructs a multi tiered basic public service system for vision health management of primary and secondary school students. It further proposes the use of an intelligent management platform and a "three tier color coded early warning" mechanism (red, yellow, blue) to prevent and control student myopia at the source, providing schools with replicable and operable implementation pathways to enhance the level of student vision health management.

ObjectiveTo address the key challenges in breast electrical impedance tomography (EIT) reconstruction including the severe ill‑posedness of the inverse problem, limited measurement information from the electrodes, low contrast of small lesions, as well as the common drawbacks of traditional reconstruction methods such as boundary blurring, low spatial resolution, and weak small-target detection capability, and to overcome the limitations of existing deep learning approaches namely low training efficiency, insufficient physical consistency, and substantial information loss during the low‑to‑high dimensional mapping process, this study proposes a multi‑scale attention and dynamic sparse mask fusion network (MADS‑Net). The proposed method aims to enhance high‑resolution reconstruction accuracy, small-lesion detectability, noise robustness, and computational efficiency in breast‑screening-oriented EIT, thereby providing a feasible solution for the application of EIT in breast functional imaging. MethodsThe proposed network adopts a strategy of “locate first, reconstruct later”. It realizes adaptive allocation of computational resources through a dynamic mask module, alleviates information loss in the low-to-high dimensional mapping process via a multi-scale attention mechanism, and embeds physical constraints to improve the physical rationality of reconstruction results. The performance of the proposed method was validated through systematic numerical simulations, agar phantom experiments, and porcine tissue experiments. Comparisons were made with mainstream methods including the Gauss-Newton algorithm, convolutional neural networks, U-Net, and Deep D-Bar. The image correlation coefficient (ICC), root mean square error (RMSE), and peak signal-to-noise ratio (PSNR) were adopted as quantitative evaluation metrics. Furthermore, lesion‑level evaluation metrics were introduced to assess the capability of different methods in reconstructing abnormal target boundaries and small‑sized target regions. ResultsExperimental results demonstrate that MADS-Net outperforms all comparison methods across all test scenarios. In numerical simulations, it achieves higher accuracy in target morphology and spatial location recovery as well as better background suppression. In agar phantom experiments, its detection and imaging reliability for small targets is significantly superior to the comparison methods. In porcine tissue experiments, the model maintains good generalization performance and can adapt to complex biological tissue imaging scenarios. Quantitative results show that MADS-Net achieves an average ICC of 0.917 6±0.048 8, an RMSE as low as0.060 6±0.031 6, and a PSNR of (23.893 9±2.685 3) dB. It successfully detects tiny targets as small as 5 mm within a 100 mm diameter sensing field, reduces the total training duration by 23.5% compared with the baseline method, and maintains stable reconstruction performance even under strong noise interference. ConclusionThe results demonstrate that MADS-Net achieves favorable EIT reconstruction performance in numerical simulations, agar phantom experiments, and porcine tissue experiments, even under common practical imaging challenges including low signal-to-noise ratio and blurred tissue boundaries. This study provides a methodological reference for further research on EIT in breast-screening-related scenarios and other medical functional imaging tasks, and also offers a low-cost, radiation-free alternative for breast lesion screening that aligns with the clinical demand for non-invasive diagnostic tools. It should be noted that real clinical breast data have not yet been used for validation in this study, and the clinical applicability of the proposed method requires further investigation. Future work will prioritize collecting multi-center real clinical breast datasets, and further optimize the model’s generalization ability to adapt to individual differences in breast tissue composition to promote its clinical application.