Amid the global wave of digital economy, China's medical artificial intelligence applications are rapidly advancing through technological innovation and policy support, while facing multifaceted evaluation and regulatory challenges. The dynamic algorithm evolution undermines the consistency of assessment criteria, multimodal systems lack unified evaluation metrics, and conflicts persist between data sharing and privacy protection. To address these issues, the China National Health Development Research Center has established a value assessment framework for artificial intelligence medical technologies, formulated the country's first technical guideline for clinical evaluation, and validated their practicality through scenario-based pilot studies. Furthermore, this paper proposes introducing a "regulatory sandbox" model to test technical compliance in controlled environments, thereby balancing innovation incentives with risk governance.
Artificial Intelligence/legislation & jurisprudence* ; China ; Humans ; Algorithms

Disinfection by-product(DBPs)are contaminants generated during drinking water treatment processes.Despite their low concentration level,these compounds exhibit high toxicity,posing threaten to both environmental safety and human health.Traditional DBPs analysis methods rely on chromatography/mass spectrometry techniques,which are limited by complex and time-consuming pretreatment processes,as well as expensive and non-portable instrumentation.Therefore,there is an urgent need to develop sensitive,fast,simple and low-cost in-situ detection technique and analysis instruments for DBPs.Electrochemical sensors,as a beneficial complement to the standard DBPs detection method,are expected to achieve on-site in-situ detection and remote real-time monitoring.This article provided a concise overview of regulatory indicators and standardized detection methods for DBPs,followed by an in-depth discussion of recent advancements in electrochemical detection of DBPs,focusing on two key aspects,recognition probes and analytical techniques.Finally,the current challenges and potential research directions in the field of electrochemical sensors for DBPs were summarized.

Objective:To predict mechanical complications (MC) following spinal deformity surgery for adult spine deformity (ASD) using machine learning models, identify key risk factors, and develop a visualizable tool for individualized risk assessment.Methods:Clinical and radiological data from 525 patients with ASD who underwent surgery in our hospital between January 2017 and December 2021 were collected. Patients were randomly assigned to a training set (70%) and a test set (30%) for model development. The cohort included 88 males and 437 females, with a mean age of 42.2±18.1 years. Variables included demographic data, comorbidities, local and systemic radiological parameters, paraspinal muscle fat infiltration (FI), and vertebral bone quality (VBQ) scores. Multiple machine learning algorithms: Random Forest (RF), Gaussian Naive Bayes (GNB), Light GBM, Support Vector Machine (SVM), XGBoost (XGB), and Logistic Regression (LR) were trained and evaluated. Model performance was compared using the receiver operating characteristic curve (ROC) and precision-recall curve (PRC). SHAP (Shapley Additive Explanations) was used to rank risk factors, while LIME (Local Interpretable Model-Agnostic Explanations) was applied to visualize MC risk in individual cases.Results:Of the 525 patients, 135 (25.7%) developed postoperative MC. Among these, 80 (59.3%) experienced proximal junction kyphosis or failure (PJK/PJF), 7 (5.2%) had distal junction kyphosis or failure (DJK/DJF), 28 (20.7%) sustained rod fractures, and 29 (21.5%) showed significant loss of correction. In the validation cohort, the RF model achieved the highest area under the curve (AUC=0.80), followed by GNB (0.77), XGB (0.76), LR (0.74), LightGBM (0.73), and SVM (0.66). The RF model also demonstrated the best PRC value (0.58), highest sensitivity (0.65), and lowest Brier score (0.20). GNB, Light GBM, and LR models achieved the highest accuracy (0.78 each), while LightGBM exhibited the highest specificity (0.93). SHAP analysis identified higher preoperative VBQ scores, larger T 1 pelvic angle (TPA), and higher paraspinal muscle FI as the main risk factors for MC. Based on the RF model, a LIME-based tool was successfully constructed for individualized MC risk estimation. Conclusion:The RF model demonstrated the best overall predictive performance for MC. A machine learning-based prediction model has the potential to provide valuable guidance for surgical decision-making in ASD patients.

Objective:To predict mechanical complications (MC) following spinal deformity surgery for adult spine deformity (ASD) using machine learning models, identify key risk factors, and develop a visualizable tool for individualized risk assessment.Methods:Clinical and radiological data from 525 patients with ASD who underwent surgery in our hospital between January 2017 and December 2021 were collected. Patients were randomly assigned to a training set (70%) and a test set (30%) for model development. The cohort included 88 males and 437 females, with a mean age of 42.2±18.1 years. Variables included demographic data, comorbidities, local and systemic radiological parameters, paraspinal muscle fat infiltration (FI), and vertebral bone quality (VBQ) scores. Multiple machine learning algorithms: Random Forest (RF), Gaussian Naive Bayes (GNB), Light GBM, Support Vector Machine (SVM), XGBoost (XGB), and Logistic Regression (LR) were trained and evaluated. Model performance was compared using the receiver operating characteristic curve (ROC) and precision-recall curve (PRC). SHAP (Shapley Additive Explanations) was used to rank risk factors, while LIME (Local Interpretable Model-Agnostic Explanations) was applied to visualize MC risk in individual cases.Results:Of the 525 patients, 135 (25.7%) developed postoperative MC. Among these, 80 (59.3%) experienced proximal junction kyphosis or failure (PJK/PJF), 7 (5.2%) had distal junction kyphosis or failure (DJK/DJF), 28 (20.7%) sustained rod fractures, and 29 (21.5%) showed significant loss of correction. In the validation cohort, the RF model achieved the highest area under the curve (AUC=0.80), followed by GNB (0.77), XGB (0.76), LR (0.74), LightGBM (0.73), and SVM (0.66). The RF model also demonstrated the best PRC value (0.58), highest sensitivity (0.65), and lowest Brier score (0.20). GNB, Light GBM, and LR models achieved the highest accuracy (0.78 each), while LightGBM exhibited the highest specificity (0.93). SHAP analysis identified higher preoperative VBQ scores, larger T 1 pelvic angle (TPA), and higher paraspinal muscle FI as the main risk factors for MC. Based on the RF model, a LIME-based tool was successfully constructed for individualized MC risk estimation. Conclusion:The RF model demonstrated the best overall predictive performance for MC. A machine learning-based prediction model has the potential to provide valuable guidance for surgical decision-making in ASD patients.

Objective:To summarize and analyze the composition characteristics and problems of basic medical insurance policies for traditional Chinese medicine in various provinces of China,providing reference for optimizing and improving subsequent basic medical insurance policies for traditional Chinese medicine.Methods:Based on the perspective of policy instrument,combined with two dimensions of policy instrument types and policy development process,the content analysis method is used to quantitatively analyze the content of the basic medical insurance policies for traditional Chinese medicine released at the provincial level from 2011 to 2023.Results:The 93 included policy documents were coded and sorted,with a cumulative total of 487 codes.From the perspective of policy instrument dimensions,subcategories of policy instruments involve diverse themes,but there are differences in the level of attention paid to each policy tool.From the perspective of policy development process,each link also presents a discrete trend,indicating a dominant feature of policy planning and implementation.Conclusion:To improve the basic medical insurance policy system of traditional Chinese medicine in China,it is necessary to optimize the combination of policy instrument and construct a coordinated and balanced policy instrument framework;Overall planning of the development process of traditional Chinese medicine medical insurance policies,highlighting the unique advantages of traditional Chinese medicine;Emphasize policy synergy between dimensions and strengthen the implementation of traditional Chinese medicine medical insurance policies.

Adaptive immunity is a critical component of the human immune system, playing an essential role in identifying antigens and orchestrating a tailored immune response. This review delves into the significant strides made in the development of epitope prediction tools, their integration into vaccine design, and their pivotal role in enhancing immunotherapy strategies. The review emphasizes the transformative potential of these tools in refining our understanding and application of immune responses. Adaptive immunity distinguishes itself from innate immunity by its ability to recognize specific antigens and remember past infections, leading to quicker and more effective responses upon subsequent exposures. This facet of immunity involves complex interactions between various cell types, primarily B cells and T cells, which recognize distinct epitopes presented by antigens. Epitopes are small sequences or configurations on antigens that are recognized by the immune receptors on B cells and T cells, acting as the focal points of immune recognition and response. Epitopes can be broadly classified into two types: linear (or sequential) epitopes and conformational (or discontinuous) epitopes. Linear epitopes consist of a sequence of amino acids in a protein that are recognized by B cells and T cells in their primary structure form. Conformational epitopes, on the other hand, are formed by spatially distinct amino acids that come together in the tertiary structure of the protein, often recognized by the immune system only when the protein folds into its native conformation. The role of epitopes in the immune response is critical as they are the primary triggers for the activation of B cells and T cells. When an epitope is recognized, it can stimulate B cells to produce antibodies, mobilize helper T cells to secrete cytokines, or prompt cytotoxic T cells to kill infected cells. These actions form the basis of the adaptive immune response, tailored to eliminate specific pathogens or infected cells effectively. The prediction of B cell and T cell epitopes has evolved with advances in computational biology, leading to the development of several sophisticated tools that utilize a variety of algorithms to predict the likelihood of epitope regions on antigens. Tools employing machine learning methods, such as support vector machines (SVMs), XGBoost, random forest, analyze large datasets of known epitopes to classify new sequences as potential epitopes based on their similarity to known data. Moreover, deep learning has emerged as a powerful method in epitope prediction, leveraging neural networks capable of learning high-dimensional data from vast amounts of immunological inputs to identify patterns that may not be evident to other predictive models. Deep learning models, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs) and ESM protein language model have demonstrated superior accuracy in mapping the nonlinear relationships inherent in protein structures and epitope interactions. The application of epitope prediction tools in vaccine design is transformative, enabling the development of epitope-based vaccines that can elicit targeted immune responses against specific parts of the pathogen. These vaccines, by focusing the immune response on highly specific regions of the pathogen, can offer high efficacy and reduced side effects. Similarly, in cancer immunotherapy, epitope prediction tools help identify tumor-specific antigens that can be targeted to develop personalized immunotherapeutic strategies, thereby enhancing the precision of cancer treatments. The future of epitope prediction technology appears promising, with ongoing advancements anticipated to enhance the precision and efficiency of these tools further. The integration of broader immunological data, such as patient-specific immune profiles and pathogen variability, along with advances in AI and machine learning, will likely drive the development of more adaptive, robust, and clinically relevant prediction models. This will not only improve the effectiveness of vaccines and immunotherapies but also contribute to our broader understanding of immune mechanisms, potentially leading to breakthroughs in the treatment and prevention of multiple diseases. In conclusion, the development and refinement of epitope prediction tools stand as a cornerstone in the advancement of immunological research and therapeutic design, highlighting a path toward more precise and personalized medicine. The ongoing integration of computational models with experimental immunology holds the promise of revolutionizing our approach to combating infectious diseases and cancer.

OBJECTIVE: To investigate the factors related to renal impairment in patients with diabetic kidney disease (DKD) from the perspective of integrated Chinese and Western medicine. METHODS: Totally 492 patients with DKD in 8 Chinese hospitals from October 2017 to July 2019 were included. According to Kidney Disease Improving Global Outcomes (KDIGO) staging guidelines, patients were divided into a chronic kidney disease (CKD) 1-3 group and a CKD 4-5 group. Clinical data were collected, and logistic regression was used to analyze the factors related to different CKD stages in DKD patients. RESULTS: Demographically, male was a factor related to increased CKD staging in patients with DKD (OR=3.100, P=0.002). In clinical characteristics, course of diabetes >60 months (OR=3.562, P=0.010), anemia (OR=4.176, P<0.001), hyperuricemia (OR=3.352, P<0.001), massive albuminuria (OR=4.058, P=0.002), atherosclerosis (OR=2.153, P=0.007) and blood deficiency syndrome (OR=1.945, P=0.020) were factors related to increased CKD staging in patients with DKD. CONCLUSIONS Male, course of diabetes >60 months, anemia, hyperuricemia, massive proteinuria, atherosclerosis, and blood deficiency syndrome might indicate more severe degree of renal function damage in patients with DKD. (Registration No. NCT03865914).
Humans ; Male ; Diabetes Mellitus, Type 2 ; Diabetic Nephropathies ; Hyperuricemia ; Kidney ; Proteinuria ; Renal Insufficiency, Chronic/complications*

Objective: To establish a predictive model for survival benefit of patients with intrahepatic cholangiocarcinoma (ICC) who received adjuvant chemotherapy after radical resection. Methods: The clinical and pathological data of 249 patients with ICC who underwent radical resection and adjuvant chemotherapy at 8 hospitals in China from January 2010 to December 2018 were retrospectively collected. There were 121 males and 128 females,with 88 cases>60 years old and 161 cases≤60 years old. Feature selection was performed by univariate and multivariate Cox regression analysis. Overall survival time and survival status were used as outcome indicators,then target clinical features were selected. Patients were stratified into high-risk group and low-risk group,survival differences between the two groups were analyzed. Using the selected clinical features, the traditional CoxPH model and deep learning DeepSurv survival prediction model were constructed, and the performance of the models were evaluated according to concordance index(C-index). Results: Portal vein invasion, carcinoembryonic antigen>5 μg/L,abnormal lymphocyte count, low grade tumor pathological differentiation and positive lymph nodes>0 were independent adverse prognostic factors for overall survival in 249 patients with adjuvant chemotherapy after radical resection (all P<0.05). The survival benefit of adjuvant chemotherapy in the high-risk group was significantly lower than that in the low-risk group (P<0.05). Using the above five features, the traditional CoxPH model and the deep learning DeepSurv survival prediction model were constructed. The C-index values of the training set were 0.687 and 0.770, and the C-index values of the test set were 0.606 and 0.763,respectively. Conclusion: Compared with the traditional Cox model, the DeepSurv model can more accurately predict the survival probability of patients with ICC undergoing adjuvant chemotherapy at a certain time point, and more accurately judge the survival benefit of adjuvant chemotherapy.