Deep brain stimulation (DBS) can provide an important treatment option for refractory diseases such as Parkinson’s disease, dystonia and epilepsy through targeted modulation of neural circuits, helping to improve patients’ quality of life. With the advancement of brain science research and innovations in engineering technology, the clinical indications for DBS have been continuously expanded, and its technical architecture has also been iteratively upgraded. This paper systematically reviewed the development history of DBS, from its early technical embryonic form to current intelligent technologies including directional electrodes and adaptive closed-loop stimulation. It focused on the analysis of regulatory frameworks and clinical guidelines in major regions such as the European Union, the United States and China, as well as the differences in product evaluation requirements between China and the United States. Meanwhile, it compared and analyzed the technical characteristics and clinical application status of domestic and imported DBS products. Centering on the regulatory challenges faced by this field, the paper proposed approaches to improve scientific supervision and policies in line with China’s national conditions, so as to provide theoretical references and practical strategies for the standardized and internationalized development of China’s DBS field.

Objective To solve the imaging quality degradation problem of high-resolution flat-panel detectors (HR-FPDs) in X-ray low-dose acquisition mode, to explore a self-supervised domain adaptation method to significantly improve the imaging performance of HR-FPDs in actual low-dose imaging scenarios. Methods Firstly, a large-scale animal joint imaging dataset was collected using HR-FPDs, and the source domain model was pre-trained based on this dataset. By combining iterative knowledge transfer and style generalization learning methods, the source domain model was transferred and trained to the target domain with only a small amount of unlabeled human joint data. Finally, through experiments, the improvement effect of the proposed self-supervised domain adaptation learning method on the image quality of HR-FPDs’ two-dimensional projection imaging and cone-beam CT threedimensional reconstruction under low-dose imaging conditions was verified. Results The experimental results showed that after noise reduction processing of low-dose data collected by large-size HR-FPDs using the self-supervised domain adaptation method, the imaging performance was significantly improved. In two-dimensional projection imaging, the signal-to-noise ratio of the projection images under the low-dose exposure mode increased by 6.66 and 9.95 dB on the two test sets, with statistically significant differences (P<0.001); the reconstruction quality of cone-beam CT was also optimized: the noise and artifacts in the reconstructed images were significantly suppressed, and the CT values in low-density areas decreased, with statistically significant differences (P<0.001). Conclusion The proposed self-supervised domain adaptation learning method can effectively improve the imaging performance of HR-FPDs under low-dose acquisition conditions, thereby expanding its applicability in more X-ray low-dose application scenarios.

Objective To enhance the objectivity and accuracy of traditional Chinese medicine syndrome differentiation and address the issues of reliance on expert experience and strong subjectivity, aiming to construct an intelligent diagnosis and treatment model based on multimodal clinical data such as tongue images, pulse images, and symptom texts. Methods Firstly, an improved convolutional recurrent neural network (CRNN) was designed to extract spatial features of tongue images, process temporal features of pulse patterns, and sequence of symptom texts; a spatial reconstruction unit was introduced to adaptively capture irregular pathological features such as fissures and bruises on tongue images through dynamic deformable convolution kernels; a channel reconstruction unit was designed to select key semantic channels by combining depth separable convolution and compressionincentive mechanisms; a three-layer cross-modal attention mechanism was adopted to dynamically integrate multimodal features, and a three-level syndrome priority matching algorithm was integrated to simulate the priority traditional Chinese medicine syndrome differentiation process based on main symptoms. Results The intelligent diagnosis and treatment model based on traditional Chinese medicine clinical medical data and the improved CRNN model had good learning ability and generalization ability. In the results of syndrome differentiation for qi-yin deficiency and qi deficiency with blood stasis, the accuracy of the improved CRNN model was the highest, reaching 88.69% and 97.62% respectively; in the precision-recall curve and receiver operating characteristic (ROC) curve, the areas under the curves for swollen tongue and punctate tongue were the highest, being 0.9737 and 0.9713 respectively. The ROC curve and precision-recall curve verified the balance of model performance. Conclusion This model can effectively solve the problems of multimodal feature extraction and fusion in traditional Chinese medicine, significantly improve the objectivity of syndrome differentiation, and provide a reliable tool for clinical intelligent diagnosis and treatment.

Objective To enhance the robustness and accuracy of the lung ultrasound image detection system to assist physicians in the early diagnosis of lung diseases. Methods Based on TransUNet, the self-attention mechanism and convolutional operation hybrid model were introduced to improve the precise detection of lesions. At the same time, the EfficientNet version 2 transfer learning technology was combined to achieve accelerated convergence of the model. Results The F1 score of the optimized model based on ACmix and transfer learning (hereinafter referred to as the “optimized model”) was 0.943, while the F1 score of the basic model TransUNet was 0.891. The classification performance of the optimized model was better. The intersection over union (IoU) of the optimized model was 0.924, and that of TransUNet was 0.862. The segmentation effect of the optimized model was better. According to the segmentation effect of lung ultrasound images and the comparison results of actual cases, it could be seen that the optimized model performed well in classifying different basic signs of lung ultrasound images, demonstrating the efficiency and accuracy of the optimized model in the diagnosis of lung diseases. Conclusion The optimized model proposed in this paper can improve the efficiency of early detection and diagnosis of lung diseases and provide intelligent technical support for ultrasound image detection.

Objective To solve the problems of strong data heterogeneity and insufficient model generalization ability existing in the data analysis of medical equipment, and to establish an efficient and reliable data analysis algorithm for medical equipment. Methods Firstly, a discriminative feature space was constructed through an encoding comparison mechanism, and the model gradients of each client were regarded as relevant sources. Only the residuals between the common edge information with the server were uploaded, thereby significantly reducing communication overhead while enhancing the consistency of cross-client feature representations. Then, a classification correction mechanism was introduced, where the output logic of the local model was dynamically calibrated based on the global data distribution prior. The relative entropy constraint was used to sharpen the blurred decision boundaries due to data heterogeneity, effectively suppressing model bias. Results Experimental results showed that in the RICORD and MIMIC-CXR datasets, the classification label consistency rates of the proposed algorithm were 96.08% and 96.58%, respectively, and the agreement rates with doctor’s diagnoses were 97.52% and 96.87%, respectively, verifying its excellent clinical consistency. At the same time, in the infusion pump and ventilator data, the data leakage risk rates of the proposed algorithm were 2.04% and 1.93%, respectively, lower than those of the other three algorithms. In the data analysis scenarios of infusion pumps and ventilators, the fault detection rate, data integrity rate, equipment data utilization improvement rate, and analysis time of the proposed algorithm were 98.12%, 98.23%, 15.07%, 131.2 ms, and 98.34%, 98.42%, 16.01%, 130.5 ms, respectively, superior to those of the other three comparison algorithms, demonstrating outstanding performance. Conclusion This research algorithm can provide reliable technical support for privacy security and diagnostic accuracy in smart healthcare scenarios.

Objective To evaluate the safety and partial efficacy of high frequency-irreversible electroporation (HF-IRE) in ablating rats with orthotopic brain tumors, and to provide key parameter basis for clinical translation. Methods A rat model of in situ brain tumors was established, and HF-IRE treatment was performed using three gradient electric field intensities of 900/1200/1500 V/cm. The tumor volume was dynamically monitored by magnetic resonance imaging, and the pathological damage and cell apoptosis of brain tissue were evaluated by hematoxylin-eosin staining (H&E) and TdT-mediated dUTP nick-end labeling (TUNEL) methods to systematically assess safety and efficacy. Results During the operation, the HFIRE system operated stably with no device-related abnormalities. In the HFIRE group, red blood cell count and platelet levels increased postoperatively, whereas those in the control group decreased on postoperative day 14. There was no statistically significant difference in aspartate aminotransferase between the two groups (P>0.05); alanine aminotransferase and alkaline phosphatase levels increased postoperatively (P<0.05) and returned to baseline by day 14. Both groups exhibited acute inflammatory responses, with no statistically significant difference in pathological intensity (P>0.05). On postoperative day 14, the apoptosis score in the 1500 V/cm group was higher than those in the 900 and 1200 V/cm groups (F=7.038, P=0.046). Nissl staining showed minor neuronal injury around the electrode needle puncture channels in both groups at postoperative day 3; neuronal injury in the HFIRE groups was slightly higher at day 3 than at day 14. Scores at day 14 indicated high safety for the 900 and 1200 V/cm field strengths, while 1500 V/cm might induce mild neuronal injury within the ablation zone. Conclusion 1200 V/cm is the optimal balance parameter for HF-IRE treatment of brain tumors, which can ensure a significant tumor inhibition effect (>90%) while minimizing the adverse reactions and nerve injury risks caused by HF-IRE.

Objective To verify the system stability of the cadmium zinc telluride cardiacdedicated scanner through retrospective analysis of its daily quality control data, and to explore its influencing factors and corresponding solutions. Methods According to the user manual, daily quality control data of the DSPECT cardiacdedicated scanner and temperature and humidity data of the equipment room from 2017 to 2023 were collected. Nonparametric Spearman’s rank correlation analysis and multiple linear regression models were adopted to investigate the influences of temperature and humidity in the equipment room and the activity of the quality control source on the quality control indicators. Meanwhile, the nonparametric Mann-Whitney U test was utilized to evaluate the differences before and after the replacement of the quality control source. Results All quality control results were qualified within the 6 year period, during which three quality control rod sources were used. The results of nonparametric Spearman’s rank correlation analysis showed that there were weak correlations between the temperature in the equipment room housing the cardiacdedicated scanner and five quality control indicators in the daily quality control results, including the local uniformity index, global uniformity index, effective field of view and energy resolution, and probe registration index (rs=-0.157, -0.211, 0.083, 0.217, 0.117, P<0.001). The humidity in the equipment room presented weak correlations with partial quality control indicators (local uniformity index: rs=-0.073, P<0.001; global uniformity index: rs=-0.123, P<0.001; effective field of view: rs=0.070, P<0.001; energy resolution: rs=-0.055, P=0.029), while no significant correlation was detected between humidity and the probe registration index (rs=-0.041, P=0.104). The activity variations of two quality control rod sources exhibited significant correlations with all five quality control indicators (rod source 1 activity: rs=0.352, 0.399, -0.157, 0.257, 0.440, P<0.001; rod source 2 activity: rs=0.074, 0.225, 0.718, 0.777, 0.180, P<0.001). The results of multiple linear regression analysis indicated that the temperature in the equipment room (β=0.122, 0.088, P<0.001) and the activity of the rod source (β=0.614, 0.742, P<0.001) could significantly and positively predict the effective field of view and energy resolution. The humidity in the equipment room (β=0.019, P=0.478) failed to significantly predict the effective field of view, but could significantly and negatively predict the energy resolution (β=-0.203, P<0.001). The nonparametric dichotomous Mann-Whitney U test revealed that the two replacements of the rod source exerted significant effects on all five quality control indicators (P<0.05). Conclusion The DSPECT cardiacdedicated scanner maintains stable system performance throughout its 6-year service period, and its stability in clinical application can be guaranteed under strict compliance with operational specifications. It is recommended that attention be paid to environmental changes in the equipment room and the service status of the rod source during operation to ensure the normal functioning of the scanner.

Objective To conduct a comparative analysis of the results of composite fields and large fields in Portal Dosimetry (PD) verification for fixed-field intensity-modulated radiation therapy plans of cervical cancer, and to provide reference and guidance for medical physicists in performing dosimetry verification work. Methods A total of 21 fixed-field intensity-modulated radiation therapy plans for patients with cervical cancer were selected as the research objects, and each treatment plan contained 7 large fields, with each large field composed of a pair of subfields. Two types of PD verification plans, including only subfields and only large fields respectively, were established using the Eclipse treatment planning system, and the verification plans were executed using a Clinac iX linear accelerator. The Gamma passing rates (G-value) of large fields were directly obtained by the PD software module; G-value of the corresponding composite fields were derived from a pair of subfields using the composite field menu tool. The Gamma analysis was performed with the parameter settings of 3 mm/3% and a threshold of 10%. Statistical analysis and comparative study were conducted on the G-values of composite fields, large fields and subfields. The rank-sum test was used to compare the two groups of G-value data of composite fields and large fields. Results A total of 147 G-values for composite fields and 147 G-values for large fields were obtained, with median values of 99.2% and 98.3% respectively. The median value of the ΔG-value between composite fields and large fields was 0.9%. The median G-values of subfield 1, subfield 2 and all subfields were 99.7%, 99.9% and 99.8% respectively. The G-values of the two types of subfields and all subfields were generally higher than the corresponding values of composite fields and large fields. The results of the rank-sum test for pairwise comparison of G-values among the three field types (composite fields, large fields, subfields) showed that the differences were statistically significant (P<0.05). Conclusion In PD verification for cervical cancer intensity-modulated radiation therapy plans, the G-value of composite fields is generally higher than that of large fields, and composite fields can be considered to achieve favorable dosimetry verification results.

Objective To establish a multimodal image-based automatic segmentation model for the clinical target volume (CTV) and organs at risk (OAR) in cervical cancer radiotherapy, and to construct and validate this model based on deep learning neural networks using simulation CT images and corresponding magnetic resonance images of patients receiving cervical cancer radiotherapy. Methods A total of 150 cases of simulation CT images and corresponding T2-weighted magnetic resonance images from patients who underwent cervical cancer radiotherapy at Northern Jiangsu People’s Hospital from January 2022 to September 2023 were retrospectively collected and preprocessed. These cases were randomly divided into a training set of 100 cases, a validation set of 25 cases and a test set of 25 cases. The multimodal images were simultaneously input into an automatic segmentation model constructed based on a deep learning neural network for training and validation, and automatic segmentation was then performed on the test set. Taking the manual segmentation results of physicians as the “gold standard”, the accuracy of the automatic segmentation model for cervical cancer CTV and OAR (small intestine, bladder, rectum, left and right kidneys, left and right femoral heads) was calculated, and the time consumed for segmentation was recorded. Results The Dice similarity coefficient values of the automatic segmentation model for CTV and OAR (small intestine, bladder, rectum, left and right kidneys, left and right femoral heads) were 0.87±0.03, 0.79±0.04, 0.95±0.04, 0.88±0.04, 0.96±0.02, 0.96±0.03, 0.92±0.03 and 0.93±0.03, respectively. The 95% Hausdorff distance values were (5.12±1.45), (22.37±15.68), (1.27±0.31), (5.45±1.56), (1.15±0.21), (1.22±0.25), (4.51±2.38) and (4.56±2.77) mm, respectively. The Jaccard index values were 0.88±0.05, 0.83±0.04, 0.97±0.02, 0.91±0.04, 0.97±0.02, 0.97±0.03, 0.98±0.02 and 0.98±0.01, respectively. The time consumed for segmentation was (0.53±0.09), (0.36±0.06), (0.08±0.02), (0.09±0.03), (0.03±0.01), (0.03±0.01), (0.04±0.01) and (0.04±0.01) min, respectively, and the total time consumed to completely segment one case was (1.20±0.24) min. All differences were statistically significant (P<0.05). Conclusion The automatic segmentation model based on multimodal images established in this study can accurately segment the CTV and OAR for cervical cancer radiotherapy automatically, providing a certain reference for clinicians in target volume and organ segmentation and saving a substantial amount of clinical time.

Objective To propose an optimization mechanism based on hierarchical chain-cloud architecture and smart contracts, as to solve the problems of insufficient data security, weak protection of patient privacy and low efficiency of cross-institutional business collaboration in traditional hospital business systems. Methods This study combines hierarchical chain-cloud architecture and smart contracts, integrating Hyperledger Fabric blockchain with MongoDB cloud database to automate medical record authorization and claims review. The blockchain stores medical record indexes and authorization logs, while medical data is stored in the cloud database, employing a layered storage mechanism to enhance data security and access efficiency. Results The average access time of the optimized system was (0.49±0.07) s, and the illegal access detection rate reached to 98.67%. The response times for medical record authorization and claims review were (0.27±0.04) and (0.85±0.06) s respectively, and the traceability reached to 99.92%. The crash rate of the optimized system in high-concurrency scenarios is controlled at 4.63%, and the recovery time does not exceed 2.28 s. Conclusion The research verified the advantages of the proposed optimization mechanism in terms of efficiency, security and scalability, providing a practical reference for the construction of the new generation of medical information systems.

Objective To address inefficiencies in traditional medical consumables management, such as poor information tracking and inadequate supervision, this researd developed an supply processing distribution management platform based on the unique device identifier (UDI) system, enhance the efficiency of consumable management and improve the safety and quality of healthcare. Methods Core requirements were identified through user research. The platform employs a UDI-centered architecture with three layers: external data interfacing, internal UDI-based materials management, and internal system integration. A layered technical framework and 6 functional modules enable closed-loop, full-process management. Performance metrics from 20 clinical departments were compared pre- and post-implementation. Results After implementation, weekly average time consumption for requisitioning, collection, inventory counting, and billing decreased by 49.6%, 46.2%, 46.3%, and 48.0%, respectively (P<0.001). The billing error rate for high-value consumables dropped from 3.3% to 1.3% (P<0.001), indicating significant improvements in efficiency and accuracy. Conclusion By deeply integrating technological innovation with workflow optimization, this platform resolves key challenges in traditional management and provides a replicable solution for standardizing and digitizing medical supplies management.

Objective To address the issues of insufficient concurrent processing capacity of traditional electronic medical record systems, as well as inefficiency and security flaws in data access control methods. Methods The research designed an electronic medical record system based on a cloud platform architecture and employed a ciphertext-policy attribute-based encryption algorithm to optimize data access. The effectiveness of the system was evaluated by testing the response time and throughput of the system under different concurrent scenarios, as well as the key generation time, encryption and decryption durations of the ciphertext-policy attribute-based encryption algorithm with different numbers of attributes. Results The system had a response time of less than 1200 ms and a throughput of 2700 times per second under the 5000-concurrent scenario. At 12 K concurrency, the maximum throughput was 3924, with all response times being less than 3000 ms. For the ciphertext-policy attribute-based encryption algorithm, the key generation time is only 1247 ms when the number of attributes was 25, and the encryption and decryption durations were only 824 and 901 ms respectively when the number of attributes was 20. Conclusion This indicates that the cloud platform-based electronic medical record system and data access optimization method designed in the research have strong feasibility and practical application potential, effectively solving the drawbacks of traditional systems, providing strong support for the efficient and secure operation of electronic medical record systems, and being of great significance to the informatization construction of the medical industry.

Objective To explore the application path and practical effect of data intelligence technology in the whole process of hospital consumables management under the background of the implementation of the volume-based procurement policy, and to provide new ideas for hospitals to achieve refined management. Methods Taking China-Japan Friendship Hospital as the research object, a data intelligent volume-based management system integrating big data, robotic process automation and intelligent set algorithm was constructed. The system adopted a multi-layer distributed architecture to connect with the hospital’s existing information systems, realizing the whole-process management of consumables. Statistical analysis was used to compare the management efficiency, number of successful statistics and statistical time before and after the application of the system. Results After the application of the data intelligent volume-based management system, the efficiency of hospital consumables management was significantly improved: the routine allocation time was shortened to less than 1 minute, the overall efficiency was increased by more than 80%, and the error rate was reduced from 3.2% to 1.1%. The number of successful statistics increased significantly in each month compared with manual statistics (P<0.05), the statistical time was reduced from several hours to about 0.1 hours, and the efficiency was increased by hundreds of times. In terms of cost control, the average procurement cost of hospital consumables decreased by 12.6%, the inventory turnover days were shortened from 45 days to 28 days, and the resource waste rate was reduced by 15%-20%. The overall satisfaction of clinical departments with the system increased from 78% to 93%, among which the satisfaction with the timeliness of consumables supply reached 95%. Conclusion The volume-based consumables management system constructed based on data intelligence technology has effectively improved the hospital’s operational response capacity and resource allocation efficiency under the volume-based procurement policy, and has the advantages of traceability, high efficiency and low risk. It is a key grasp for realizing the transformation of modern hospital consumables management and has a strong prospect for popularization and application.

Objective To explore the application effect of the integrated management combined with reprocessing risk control model for rigid endoscopic instruments in the Central Sterile Supply Department, and to provide practical evidence for improving the management quality of rigid endoscopic instruments. Methods A total of 800 rigid endoscopic instruments retrieved by the Central Sterile Supply Department of the Fifth Affiliated Hospital of Anhui Medical University from July to November 2024 were selected as the control group. Another 895 rigid endoscopic instruments recycled from December 2024 to April 2025 were selected as the observation group, which adopted the integrated management combined with reprocessing risk control model on the basis of routine management. The disinfection management status, incidence of adverse events, management quality score and surgeons’ satisfaction were compared between the two groups. Results After the implementation of the integrated management combined with reprocessing risk control model, the cleaning qualification rate, packaging qualification rate and distribution qualification rate of the observation group reached 99.44%, 98.99% and 99.55% respectively, which were significantly higher than the 96.25%, 96.13% and 97.75% of the control group. The sterilization reprocessing rate of the observation group was 2.46%, significantly lower than the 6.00% of the control group. The overall incidence of adverse events in the observation group was 2.12%, significantly lower than the 7.38% of the control group. The excellent rate of management quality in the observation group was 62.12%, significantly higher than the 14.00% of the control group (P<0.05). After the implementation of the model, the scores of surgeons’ satisfaction in all dimensions ranged from 95.93 to 96.83, which were significantly higher than the scores ranging from 86.80 to 88.83 before implementation (P<0.05). Conclusion The integrated management combined with reprocessing risk control model can significantly improve the management quality and level of rigid endoscopic instruments, which is worthy of clinical promotion and application.

Objective To construct a supply-processing-distribution (SPD) management strategy for operating room consumables from the perspective of medical-engineering integration, address the problems of ambiguous positioning, unclear rights and responsibilities, and non-standard processes in the application of the SPD model, and provide practical references for medical institutions to optimize the refined management of operating room consumables. Methods Taking the Affiliated Hospital of Nantong University as a case study, under the guidance of the medical-engineering integration concept and SPD lean management concept, an SPD management strategy for operating room consumables was established and put into practice. The management effect was evaluated by comparing the indexes including the management efficiency of single-operation consumables, the abnormal status of barcode-controlled consumables, and the satisfaction of medical staff in the operating room before and after the application of the SPD model. Results After the application of the SPD model, the management efficiency of operating room consumables was significantly improved: the time for consumables preparation, delivery, response to temporary requisition, and return to warehouse for a single operation was shortened by 66.25%, 42.42%, 74.06%, and 61.88% respectively (P<0.05); the loss rate, undercharging rate, and overcharging rate of barcode-controlled consumables were significantly reduced (P<0.05); the satisfaction of medical staff was notably elevated, with the satisfaction rate of physicians increasing from 56.67% to 90.00% and that of nurses rising from 40.00% to 93.33% (P<0.05). Conclusion The SPD management strategy based on the medical-engineering perspective can significantly optimize the management process of operating room consumables, reduce operational costs, improve management efficiency, help medical institutions improve quality and efficiency, and thus has extensive promotion value.

Objective To evaluate the technological maturity of each core sub-technology of invasive ventilators, and to provide a decision-making reference for medical institution managers, equipment purchasers and clinical engineers in the selection, procurement and maintenance of invasive ventilators. Methods Based on the technology evolution theory, the technological maturity of each core sub-technology of invasive ventilators was evaluated by combining the Delphi expert consultation method, technology readiness level (TRL) scale method, literature analysis method and expert scoring method. Results Two rounds of Delphi expert consultations were conducted among 17 multidisciplinary experts in this study, which verified the reliability of the consultation results and a high consistency of expert opinions. A total of 40 initial sub-technologies of invasive ventilators were integrated into 36 ones (25 core sub-technologies), which were divided into 7 major modules and two categories of standard and optional configurations. The maturity evaluation was completed by fitting the S-curve with the Logistic model, and the maturity stages of each core sub-technology were clarified: the drive and control technologies were mostly in the mature/declining stage; the flow detection and oxygen sensing technologies were in the transition from the growth stage to the mature stage; most of the conventional and highfrequency ventilation technologies had entered the mature stage; the conventional ventilation monitoring technologies featured high maturity; the safety and expansion technologies were mostly in the embryonic/infant stage. Combined with the TRL classification, 15 typical core sub-technologies were screened out, with their product characteristics, application scenarios and development trends showing significant differences. Conclusion The technological maturity evaluation based on the technology evolution theory and the corresponding S-curve model can scientifically and intuitively reflect the current status and trends of the technology life cycle of each core sub-technology of invasive ventilators. The evaluation results can provide a practical technical reference for relevant decisionmakers in the whole life cycle management of invasive ventilators.

Objective To explore the intelligent management index parameters corresponding to the key links in the whole-life-cycle management of emergency life support equipment, and determine the core index parameters for the intelligent whole-life-cycle management of such equipment. Methods A combination of literature review, field research, and brainstorming was used to construct an intelligent management model for emergency life support equipment. The Delphi method and questionnaire survey were employed to identify the key index parameters in the model, and the analytic hierarchy process was applied to determine the weight of each index, ultimately forming a core index parameter system. Results A total of 22 valid questionnaires were recovered (effective recovery rate of 100%), with an expert authority coefficient (Cr) of 0.93. The Cronbach’s α coefficients for indicator familiarity and importance were 0.970 and 0.827, respectively, and the Kendall’s concordance coefficients were 0.068 and 0.146 (P<0.05), indicating good data reliability and consistent expert evaluations. A core index system consisting of 5 first-level indicators (including procurement and tender management) and 32 second-level indicators was finally established. Verification via the analytic hierarchy process showed that the consistency of the judgment matrix was satisfactory. Weight ranking revealed that tender parameters (0.085), equipment metering plan (0.073), and equipment maintenance management (0.073) were the core key indicators, and 5 indicators including “planning and demonstration” received high ratings in both expert familiarity and importance. Conclusion The core index system for intelligent whole-life-cycle management constructed in this study is scientific and reliable, providing a strong reference for peers engaged in the intelligent management of emergency life support equipment. Future research should optimize the model in realworld application scenarios to further promote the process of intelligent management of emergency life support equipment.

Objective To construct a scientific and universal evaluation index system against the problems of discretized performance evaluation and weak quality supervision caused by the lack of a standardized evaluation system for electroacupuncture therapeutic apparatus at present, so as to provide theoretical support for the construction of industrial standardization. Methods The combined research methods of Delphi method and analytic hierarchy process were adopted. The evaluation index framework was initially determined and the Delphi questionnaire was designed through literature retrieval, team brainstorming and expert consultation. A pairwise comparison matrix was constructed by analytic hierarchy process, and the 1~9 scale method was used to quantify the relative importance of indicators. The weight of each level of indicators was obtained through geometric mean and normalization calculation, and the consistency test was conducted. Results Two rounds of expert consultation were conducted in this study, involving 15 multidisciplinary experts with an active coefficient of 93.3% and an authority coefficient of 0.867, and the experts’ opinions showed good consistency (the Kendall’s W value of each level of indicators ranged from 0.509 to 0.757, P<0.05). An evaluation index system for electroacupuncture therapeutic apparatus containing 3 first-level, 7 second-level and 26 third-level indicators was finally constructed. The consistency ratio (CR) of all judgment matrices was less than 0.1, indicating a reasonable weight distribution. Among them, “equipment performance” had the highest weight (0.6738), while its subordinate second-level indicator “safety performance” (0.4668) and third-level indicator “electrical safety hazards” (0.6144) were the core indicators of each level. This system was applied to evaluate 286 pieces of equipment covering 13 models from 8 manufacturers, with the comprehensive scores ranging from 58.02 to 95.37. Significant differences were found in the performance and safety of equipment of different brands, demonstrating good discrimination and reliability of the system. Conclusion The evaluation index system for electroacupuncture therapeutic apparatus formed in this study has strong scientificity and wide applicability, and can provide a standardized basis for the performance evaluation, product research and development, and quality supervision of electroacupuncture therapeutic apparatus.

Objective To explore the current applications, research hotspots, and future trends of functional magnetic resonance imaging (fMRI) in migraine research using bibliometric methods. Methods A comprehensive search was conducted for Chinese and English literature on fMRI studies in migraine published in the China National Knowledge Infrastructure (CNKI) and Web of Science (WOS) core collection databases up to February 28, 2025. VOSviewer 1.6.18 and CiteSpace 6.1.R6 software were employed to visualize and analyze authors, institutions, and keywords. Results A total of 163 Chinese articles and 423 English articles were included. Annual publications demonstrated a steady upward trend. Core research teams led by authors such as Yi Ren and Fanrong Liang were identified internationally. Keyword frequency and cluster analysis revealed research hotspots related to “acupuncture” “functional connectivity” “brain networks” and “resting-state”. Timeline and burst analysis highlighted emerging frontiers such as “functional connectivity” and “machine learning”. Conclusion fMRI is widely utilized in migraine research. Future efforts should focus on strengthening interdisciplinary collaboration, addressing research hotspots and frontiers, and promoting sustainable development in this field.

Objective To explore the efficacy of texture parameters from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) combined with conventional magnetic resonance imaging (MRI) signs in predicting the nature of nodular breast lesions. Methods A total of 122 female patients with breast imaging reporting and data system (BI-RADS) category 4 breast nodules, who visited Cangzhou People’s Hospital after detection of breast nodules by ultrasound physical examination from January 2022 to December 2024, were enrolled. The general data and MRI imaging findings of the patients were compared separately. The bootstrap sampling method was used to divide the patients into a training set (n=90) and a test set (n=32), and a conventional MRI imaging prediction model was established by using the random forest algorithm. Two radiologists delineated the regions of interest on DCEMRI images and extracted texture parameters. Univariate analysis was performed to screen out texture features with significant differences, which were then input into the random forest algorithm to construct a texture feature prediction model. Binary Logistic regression analysis was conducted on the significant features of conventional MRI imaging and texture parameters to build a combined prediction model, and receiver operating characteristic (ROC) curves were plotted to evaluate the diagnostic efficacy of each model. Results Compared with benign nodules, malignant nodules were more likely to present irregular shapes, lobulated or spiculated margins, heterogeneous internal signals, higher proportions of calcification and heterogeneous enhancement, as well as larger sizes, with all differences being statistically significant (P<0.05). Univariate analysis of texture features extracted from DCEMRI images of breast lesions showed that among the 13 texture features, 9 parameters including GLCM_Energy, GLCM_Entropy, GLCM_Contrast, GLCM_Correlation, GLCM_Inverse Difference Moment, GLRLM_Gray Level Nonuniformity, GLRLM_ Run Percentage, GLSZM_Large Area Emphasis and GLDM_Dependence Energy were significantly different between benign and malignant nodules (P<0.05). The area under the curve values (with 95% confidence intervals) of ROC curves for the conventional MRI imaging prediction model, texture feature prediction model and combined prediction model were 0.811 (0.753-0.828), 0.838 (0.796-0.871) and 0.893 (0.855-0.921), respectively; the accuracy rates were 82.22%, 85.60% and 94.47%, respectively; the sensitivities were 80.41%, 81.24% and 92.36%, respectively; and the specificities were 78.63%, 79.92% and 90.15%, respectively. Conclusion The combined prediction model of DCE-MRI texture parameters and conventional MRI signs can effectively predict the nature of nodular breast lesions, thereby providing a reference for clinical diagnosis and treatment.

Objective To analyze the risk factors for postoperative mortality in patients with Stanford type A aortic dissection (TAAD) and construct a nomogram-based predictive model. Methods The clinical data of 1521 patients who underwent TAAD surgery in multiple cardiovascular surgery centers in Jiangsu Province from January 1, 2019 to July 31, 2024 were collected as research subjects. According to the postoperative outcomes, the patients were divided into survival cases (n=1406) and death cases (n=115). Feature selection was performed using least absolute shrinkage and selection operator regression, recursive feature elimination and feature filtering methods. Univariate and multivariate logistic regression analyses were conducted to identify the independent risk factors for postoperative mortality. A nomogram predictive model for postoperative mortality risk in TAAD patients was constructed based on the screened independent predictors. The performance of the nomogram was evaluated using calibration curves, receiver operating characteristic (ROC) curves, precision-recall (PR) curves and Shapley additive explanations analysis. Results Feature selection and risk factor analysis revealed that patient age, length of hospital stay, initial intensive care unit stay duration, medical center, mechanical ventilation>24 h and acute kidney injury were independent risk factors for surgical mortality in TAAD patients (all P<0.05). A visualized nomogram predictive model for postoperative survival status of TAAD patients was established using the Cox method with the rms package in R software. The C-index of the model was 0.78 (95% confidence interval: 0.73-0.81, P<0.001), indicating good predictive ability. Analyses of the ROC curve, PR curve and calibration curve of the nomogram model showed that the area under the curve was 0.90, the average precision was 0.89 and the calibration degree was 0.89, suggesting that the model had excellent performance in predicting survival status. Conclusion The nomogram predictive model constructed in this study can effectively evaluate the risk of postoperative mortality in TAAD patients, providing clinicians with a tool to identify high-risk patients preoperatively and in the early postoperative period, thereby improving the efficacy of clinical diagnosis and treatment.

Lung cancer is one of the cancers with the highest morbidity and mortality rates worldwide. Compared with other malignancies, the overall 5-year survival rate of lung cancer is relatively low. Therefore, early diagnosis and precision treatment are particularly crucial for improving the prognosis of lung cancer patients. Artificial Intelligence (AI) technology, especially Deep Learning (DL) algorithms, can process large volumes of medical imaging data, thereby improving the early detection rate of lung cancer and optimizing treatment regimens. AI technology based on DL algorithms has demonstrated enormous potential in the medical field, particularly in the diagnosis and treatment of lung cancer. This paper systematically elaborates on the current application status of different types of AI algorithms in the diagnosis and treatment of lung cancer, summarizes the research progress of AI technology in the diagnosis, treatment and prognosis of lung cancer, and aims to provide a reference for the selection of early diagnostic methods and the formulation of precision treatment regimens for lung cancer.

Chronic obstructive pulmonary disease (COPD) is characterized by high incidence, high mortality and heterogeneous disease course. Early prognostic assessment is crucial for predicting disease progression and improving patients’ health outcomes. In recent years, with the continuous advancement of medical information technology, the application of Artificial intelligence (AI) in disease diagnosis and treatment has achieved remarkable progress. Machine learning, deep learning and multimodal technologies can process and analyze large volumes of complex data, enhance the efficacy of disease progression prediction and treatment, and bring new ideas and methods for COPD management. This paper reviewed the common algorithms and technologies of AI in the respiratory field, with a focus on its application in COPD prognostic assessment, including acute exacerbation risk and prognosis prediction, real-time monitoring and early warning, as well as the integration of biological feature analysis and disease health management. It is expected to provide a theoretical basis for the application of AI in COPD prognostic assessment and standardized management.

As a core medical device in medical rescue and transportation, the design and performance of medical stretchers are directly related to the treatment efficiency, safety, and comfort of the injured and patients. Although current mainstream stretchers have their own advantages in portability, terrain adaptability, or operational performance, they still have various drawbacks: foldable stretchers are prone to malfunction and difficult to sterilize; wheeled stretchers rely on flat roads and are bulky; scoop-type stretchers have limited load-bearing capacity and poor comfort; vacuum stretchers lack portability; and basket stretchers are restricted to specific application scenarios. Furthermore, the current level of intelligence of medical stretchers is low, making it difficult to meet the requirements of real-time monitoring and precise disposal in modern rescue operations. This paper systematically analyzed the type characteristics, material performance bottlenecks, and intelligent development trends of medical stretchers, focusing on the integrated application of the Internet of Things and artificial intelligence technologies in intelligent modules such as vital sign monitoring, path navigation, and remote diagnosis and treatment. It provided useful references for optimizing the structural reliability of stretchers, shortening the treatment response time, and constructing a collaborative rescue mode.

As a typical product of the in-depth integration of the Internet of Things, artificial intelligence, and healthcare fields, wearable devices have developed rapidly in recent years with the rapid advancement of sensor technology, low-power communication, and big data algorithms. By continuously and non-invasively collecting multiple physiological parameters of users, they provide important technical support for realizing individualized and continuous health management. Wearable devices make up for the deficiencies of traditional medical care in out-of-hospital monitoring and daily prevention, and lay a key foundation for constructing an intelligent health service system covering the entire life cycle. This paper reviewed the application of wearable devices in the field of intelligent medicine, elaborated on their application status in health monitoring, disease early warning, rehabilitation assistance, etc., analyzed the principles of sensor data collection and algorithms for health assessment and disease early warning, demonstrate application effectiveness through practical cases, summarize the challenges faced in terms of battery life, data accuracy, user acceptance, etc., as well as corresponding countermeasures, and prospect the future development trends, in order to promote the standardized application and research and development of wearable devices in the medical field.

Hepatobiliary surgery features high surgical risks and a high incidence of postoperative complications due to complex anatomy and individual variations. The traditional clinical model is highly dependent on physicians' experience, which is prone to deviations in tumor localization, vascular identification and real-time decision-making. In recent years, artificial intelligence technology has made breakthroughs in medical image analysis, three-dimensional visual reconstruction and clinical prediction models, bringing new opportunities for improving surgical efficiency and safety. However, its clinical translation is still restricted by diverse data sources, insufficient algorithm interpretability and the lack of large-sample and multi-center validation. This paper reviewed the latest advances of artificial intelligence technology in the key links of hepatobiliary surgery, including preoperative radiomics assessment, intraoperative augmented reality navigation and postoperative prognosis prediction. It also analyzes the existing challenges and discusses cutting-edge directions such as federated learning and human-machine collaborative decisionmaking, aiming to promote the standardized application of artificial intelligence technology, facilitate the in-depth integration of precision surgery and smart healthcare, and further improve patient prognosis.

Sarcopenia is a syndrome characterized by the loss of skeletal muscle mass, as well as the decline in muscle strength and physical function, which is associated with aging and various diseases. The prevalence of sarcopenia is significantly elevated in patients with end-stage organ failure due to underlying diseases, malnutrition and other factors. Studies have confirmed that pretransplant sarcopenia is closely associated with prolonged intensive care unit stay, increased risk of infection, delayed graft function recovery and higher mortality in recipients after transplantation. Imaging constitutes a commonly used approach for the diagnosis of sarcopenia at present. A series of imaging modalities, including dual-energy X-ray absorptiometry, computed tomography, magnetic resonance imaging, ultrasonography and artificial intelligence-assisted diagnostic techniques, can accurately assess muscle mass, thereby enabling the diagnosis of sarcopenia. This review elaborated on the principles and current clinical application status of the above-mentioned imaging modalities, and analyzed the correlation between sarcopenia evaluated by these methods and adverse posttransplant outcomes. It is intended to provide a reference for clinicians to select the most appropriate diagnostic method and improve the long-term quality of life of transplant recipients.

Hepatocellular carcinoma (HCC) is the main subtype of primary liver cancer. Owing to its insidious early symptoms and rapid progression, the 5-year overall survival rate of patients is merely 18%, making it still a major challenge in the global field of cancer therapy. In recent years, spectral computed tomography (spectral CT), as an emerging medical imaging modality, has shown a promising application prospect in oncology and exhibited enormous potential in the accurate diagnosis, therapeutic guidance and prognostic evaluation of tumor diseases. Against the backdrop of precision medicine, the in-depth integration of artificial intelligence and radiomics, combined with the multi-parameter quantitative analysis of spectral CT, has provided a novel technical support for the early screening, dynamic therapeutic efficacy evaluation and accurate prognostic prediction of HCC. This paper reviewed the research status and application value of quantitative parameters of spectral CT and radiomics in HCC, so as to provide a theoretical reference for clinicians and researchers in the field of precise diagnosis and treatment of HCC.