Artificial intelligence(AI)has emerged as a cutting-edge technology leading the future and is a key engine for China's development.In the innovation and research of medical devices,AI has provided critical support in the areas of intelligent diagnostic assistance,intelligent therapeutic assis-tance,intelligent monitoring,life support,et al.Ma-chine learning-enabled device software functions(ML-DSFs)have become an essential component of many medical devices.Recently,the United States Food and Drug Administration(FDA)released a draft guidance titled"Marketing Submission Rec-ommendations for a Predetermined Change Con-trol Plan for Artificial Intelligence/Machine Learn-ing(AI/ML)-Enabled Device Software Functions(Draft)."that aimed to provide a forward-looking approach to foster the development of ML medical devices.By supporting iterative updates through modifications,this approach ensures the continu-ous safety and effectiveness of the devices.This guidance represents the latest in regulatory direc-tion and is especially beneficial for enhancing the quality and efficiency of clinical trials for AI prod-ucts.Therefore,we plan to provide a detailed intro-duction and interpretation of the guidance,with the aim of learning from international advanced regulatory concepts and experiences to promote the development of ML-DSFs with more profound international influence.

Objective:To investigate the expression level of methyltransferase-like 3 (METTL3) in nasopharyngeal carcinoma cells CNE2 and CNE-2R, and to evaluate the effect of METTL3 on cell invasion, metastasis and radiosensitivity.Methods:Real-time reverse transcription PCR and Western blot were used to detect the expression levels of METTL3 in normal nasopharyngeal epithelial cells NP69 and nasopharyngeal carcinoma cells CNE2 and nasopharyngeal carcinoma radioresistant cells CNE-2R cells. METTL3 in CNE2 and CNE-2R cells was silenced by lentivirus-mediated RNA interference technology. The metastasis and invasion abilities of the cells were detected by the scratch assay and Transwell assay. Clonogenic assay and CCK-8 assay were employed to detect the proliferation capacity and viability of cells irradiated with different X-ray doses (0, 2, 4, 6 and 8 Gy). Apoptosis was detected by flow cytometry. Methylated RNA immunoprecipitation (Me-RIP) assay was used to detect the difference in m6A modification level of c-Jun in CNE2 and CNE-2R cells after METTL3 silencing. The transcriptional stability of c-Jun in cells after silencing METTL3 was detected by actinomycin D assay. A nude mouse xenograft model was constructed to detect the effect of METTL3 on the radiosensitivity of nasopharyngeal carcinoma in vivo. Results:Compared with NP69 cells, the expression levels of METTL3 mRNA and protein were significantly increased in CNE2 cells, and the expression level was even higher in CNE-2R cells (all P<0.01). Lentivirus-mediated RNA interference technology was used to construct a stable METTL3-silencing CNE2 and CNE-2R cell lines (both P<0.01). Scratch assay and Transwell assay showed that the metastasis and invasion abilities of CNE2 and CNE-2R cells were decreased significantly after METTL3 silencing (all P<0.05). Clonogenic assay showed that silencing METTL3 significantly reduced the number of colonies and survival fraction of CNE2 and CNE-2R cells after irradiation with different doses of X-rays (all P<0.05). CCK-8 assay showed that the proliferation ability of CNE2 and CNE-2R cells was significantly reduced by silencing METTL3 (all P<0.05). After different doses of irradiation, silencing METTL3 significantly reduced the survival fraction of CNE2 and CNE-2R cells (all P<0.05). The apoptotic rate after METTL3 silencing was higher than that in the control group at the irradiation dose of 0 and 8 Gy (all P<0.05). The Me-RIP assay showed that the m6A modification level of c-Jun in CNE2 and CNE-2R cells was significantly reduced after METTL3 silencing (both P<0.01), and the actinomycin D assay showed that transcriptional stability of c-Jun was reduced. Nude mouse xenograft experiment showed that silencing METTL3 inhibited xenograft proliferation and improved its radiosensitivity. Conclusion:METTL3 is highly expressed in nasopharyngeal carcinoma cells, and METTL3 mediates m6A modification of c-Jun to improve the transcriptional stability of c-Jun and promote the expression of c-Jun, thereby promoting the invasion and metastasis of nasopharyngeal carcinoma cells and reducing their radiosensitivity.

Artificial intelligence(AI)has emerged as a cutting-edge technology leading the future and is a key engine for China's development.In the innovation and research of medical devices,AI has provided critical support in the areas of intelligent diagnostic assistance,intelligent therapeutic assis-tance,intelligent monitoring,life support,et al.Ma-chine learning-enabled device software functions(ML-DSFs)have become an essential component of many medical devices.Recently,the United States Food and Drug Administration(FDA)released a draft guidance titled"Marketing Submission Rec-ommendations for a Predetermined Change Con-trol Plan for Artificial Intelligence/Machine Learn-ing(AI/ML)-Enabled Device Software Functions(Draft)."that aimed to provide a forward-looking approach to foster the development of ML medical devices.By supporting iterative updates through modifications,this approach ensures the continu-ous safety and effectiveness of the devices.This guidance represents the latest in regulatory direc-tion and is especially beneficial for enhancing the quality and efficiency of clinical trials for AI prod-ucts.Therefore,we plan to provide a detailed intro-duction and interpretation of the guidance,with the aim of learning from international advanced regulatory concepts and experiences to promote the development of ML-DSFs with more profound international influence.

Objective:To investigate the expression level of methyltransferase-like 3 (METTL3) in nasopharyngeal carcinoma cells CNE2 and CNE-2R, and to evaluate the effect of METTL3 on cell invasion, metastasis and radiosensitivity.Methods:Real-time reverse transcription PCR and Western blot were used to detect the expression levels of METTL3 in normal nasopharyngeal epithelial cells NP69 and nasopharyngeal carcinoma cells CNE2 and nasopharyngeal carcinoma radioresistant cells CNE-2R cells. METTL3 in CNE2 and CNE-2R cells was silenced by lentivirus-mediated RNA interference technology. The metastasis and invasion abilities of the cells were detected by the scratch assay and Transwell assay. Clonogenic assay and CCK-8 assay were employed to detect the proliferation capacity and viability of cells irradiated with different X-ray doses (0, 2, 4, 6 and 8 Gy). Apoptosis was detected by flow cytometry. Methylated RNA immunoprecipitation (Me-RIP) assay was used to detect the difference in m6A modification level of c-Jun in CNE2 and CNE-2R cells after METTL3 silencing. The transcriptional stability of c-Jun in cells after silencing METTL3 was detected by actinomycin D assay. A nude mouse xenograft model was constructed to detect the effect of METTL3 on the radiosensitivity of nasopharyngeal carcinoma in vivo. Results:Compared with NP69 cells, the expression levels of METTL3 mRNA and protein were significantly increased in CNE2 cells, and the expression level was even higher in CNE-2R cells (all P<0.01). Lentivirus-mediated RNA interference technology was used to construct a stable METTL3-silencing CNE2 and CNE-2R cell lines (both P<0.01). Scratch assay and Transwell assay showed that the metastasis and invasion abilities of CNE2 and CNE-2R cells were decreased significantly after METTL3 silencing (all P<0.05). Clonogenic assay showed that silencing METTL3 significantly reduced the number of colonies and survival fraction of CNE2 and CNE-2R cells after irradiation with different doses of X-rays (all P<0.05). CCK-8 assay showed that the proliferation ability of CNE2 and CNE-2R cells was significantly reduced by silencing METTL3 (all P<0.05). After different doses of irradiation, silencing METTL3 significantly reduced the survival fraction of CNE2 and CNE-2R cells (all P<0.05). The apoptotic rate after METTL3 silencing was higher than that in the control group at the irradiation dose of 0 and 8 Gy (all P<0.05). The Me-RIP assay showed that the m6A modification level of c-Jun in CNE2 and CNE-2R cells was significantly reduced after METTL3 silencing (both P<0.01), and the actinomycin D assay showed that transcriptional stability of c-Jun was reduced. Nude mouse xenograft experiment showed that silencing METTL3 inhibited xenograft proliferation and improved its radiosensitivity. Conclusion:METTL3 is highly expressed in nasopharyngeal carcinoma cells, and METTL3 mediates m6A modification of c-Jun to improve the transcriptional stability of c-Jun and promote the expression of c-Jun, thereby promoting the invasion and metastasis of nasopharyngeal carcinoma cells and reducing their radiosensitivity.

Objective:To evaluate the volumetric modulated arc therapy (VMAT) dose verification of cervical cancer based on γ rule and dose volume histogram (DVH) and to perform correlation analysis between the evaluation results and the dose differences.Methods:Twenty cervical cancer VMAT plans were selected and performed on TrueBeam Linac. The delivered point and surface dose was measured by FC-65 G and ArcCheck and the results were compared to those calculated by Eclipse. The dose of patients was reconstructed by 3DVH. Then, differences between the reconstructed and plan value of D mean, D 95%, D 98% and D 2% of PTV, V 20Gy of left and right femoral head, V 40Gy of rectum, D 1cm 3 of cord, D 98%, D 2% and D 50% of the 50% prescription iso-dose volume (IDV), were evaluated and 3-dimensional (3D) γ was assessed for each organ. Lastly, Pearson’s correlation coefficient was used to analyze the relationship between point dose difference, 2D γ pass-rate (γ%), γ mean and 3D γ% of each organ and the dose difference. Results:Small differences were found between the point dose measured, reconstructed and the plan value. Differences between D mean of PTV, all dose parameters of IDV and plan values were all within 3% and V 40Gy of rectum showed the largest difference. As for the 3D γ%, the maximum pass rate was found for the left and right femoral head and the maximum variance for cord D 1cm 3. There was a moderate correlation between measured and reconstructed point dose deviation and dose difference of each organ, while no significant correlation was found for 2D γ%. Strong correlation was found between 3D γ% of target and D 50% of PTV/IDV and no correlation was found for other organs. Conclusion:The performance of both γ-and DVH-based evaluation can reveal dose error for dose verification, but both of them have some limitations and should be combined in clinical practice.

Objective:To evaluate the positive rate of the bronchodilation test (BDT) in asthmatic children with normal forced expiratory volume in 1 second (FEV 1), forced vital capacity (FVC) and FEV 1/FVC, so as to improve the recognition of the importance of the BDT test in asthmatic children with normal FEV 1, FVC and FEV 1/FVC. Methods:Children aged 5-14 who were diagnosed with asthma in the outpatient clinic of Shandong Provincial Hospital Affiliated to Shandong University from September 2018 to August 2019 and willing to receive pulmonary function and BDT examinations were enrolled.Data of pulmonary function of children with normal FEV 1, FVC and FEV 1/FVC were collected to analyze the rate of positive BDT results and the status of small airway function. Results:A total of 1 631 asthmatic children with normal FEV 1, FVC and FEV 1/FVC were enrolled in this study, including 1 414 children with normal pulmonary function and 217 children with small airway dysfunction.Fifteen minutes after the bronchodilator was inhaled, 127 children (87 males and 40 females) showed positive BDT results, accounting for 7.8%.Among these children, 62 cases (28.6%) with co-existing small airway dysfunction showed positive BDT results.The improvement rate of FEV 1 was 8.0% to 11.9% in 132 cases (8.1%). The FEV 1 before bronchodilator inhalation accounted for (98.5±10.3)% of the predicted value.Fifteen minutes after terbutaline sulfate inhalation, the improvement rate was 13.5% (12.5%, 16.2%). The improvement rates of forced expiratory flow at 50% of FVC exhaled (FEF 50, r=-0.339, P<0.01), forced expiratory flow at 75% of forced vital capacity exhaled (FEF 75, r=-0.400, P<0.01), maximum mid-expiratory flow(MMEF, r=-0.375, P<0.01) were negatively correlated with their baseline values.The improvement rate of FEV 1 was not associated with its baseline value ( r=-0.128, P=0.153), but negatively correlated with the baseline value of MMEF ( r=-0.231, P<0.01). Conclusions:BDT results are positive in some asthmatic children with normal FEV 1, FVC and FEV 1/FVC.It is recommended that BDT testing should be conducted as much as possible in the diagnosis and follow-up of children with typical or atypical asthma.In this way, the diagnosis can be confirmed and the current optimal results can be obtained.Meanwhile, small airway function testing is helpful for comprehensive assessment of asthma and its control level.

Fine particulate matter (PM2.5) not only directly damages lung tissue, but also can be absorbed into blood through alveolar capillaries, which is toxic to the cardiovascular system. PM2.5 can affect lipid metabolism, endothelial function, coagulation and thrombosis through oxidative stress, inflammatory reaction, autonomic nervous dysfunction and immune regulation abnormality, so that it promote arteriosclerosis, plaque instability, and increase the morbidity and mortality of cardiovascular disease. We reviewed the effects and mechanisms of PM2.5 on arteriosclerosis, in order to provide the evidence for the studies into prevention of cardiovascular diseases caused by air pollution.

Objective:To evaluate the multi-leaf collimator (MLC) performance of TrueBeam accelerator using trajectory log files.Methods:All tests were performed 5 times under different gantry-collimator angle combination. The 1 mm picket fences were constructed by static or dynamic MLC. The control ability for small-field accuracy of accelerator was evaluated. Repeatability was assessed by MLC repeat motion. The movement performance of difference velocities along one direction and the opposite direction were evaluated via a 1 cm picket fences which slipped from -7 cm to 7 cm with a uniform velocity and stopped or immediately back at 7 cm. The MLC performance in a complex program was evaluated by a cross movement test.Results:Both the static and the dynamic picket fences yielded high accuracy. The deviation spectrums of MLC in different gantry angle were consistent, however, an absolute difference of 0.001 1 mm was found. For uniform velocity movement tests with 0°gantry, the RMSE of MLC was increased from 0.015 0 mm to 0.059 8 mm when the speed was accelerated from 5 mm/s to 25 mm/s. Similar results were obtained in non-zero gantry angle. The "overspeed" effect caused by the direction change movement of MLC was less obvious than that caused by speed changed from zero to a uniform velocity movement state. There was no significant change in speed before and after the MLC crossing. The MLC speed fluctuated around the set value, which was independent of the gantry angle.Conclusion:A method for evaluating the performance of MLC using trajectory log files is established, which can evaluate the MLC performance of TrueBeam accelerator and be used for MLC rapid quality control in clinical practice.

Objective To establish and evaluate a morning check system for linac based on electronic portal image device (EPID).Methods Delivered fluence maps of open and wedge fields at 10 cm×10 cm field size of Synergy Linac were measured by EPID.Figure features from these two images were extracted with matlab codes and analyzed to realize a quick morning check.The repeatability of dose response and mechanical setup,relationship between gray value and machine unit (MU),accuracy of output and field size test were investigated with both EPID and DailyQA3.The status of Synergy linac was monitored both by DailyQA3 and EPID for two months.Results EPID was able to test the linac consistently with a testing error of 0.50 mm,1.00 mm for field size and center,respectively.Both of the test accuracy for flatness and symmetry was 0.17％.The mechanical accuracy test and dosimetric repeatability test were also consistent.The dose response of EPID was linearly related to the linac output (R2＞0.999).EPID was highly sensitive to the change of output and radiation field size.The measurement deviations between EPID and DailyQA3 were consistent and within clinical acceptable tolerance.Conclusions EPID showed great accuracy and stability on monitoring the performance of linac.The established daily check tool based-on EPID is accurate and reliable for clinical usage.

Objective: To screen and verify the differential expression profiles of long non-coding RNAs(LncRNAs) in peripheral blood of patients with type 2 diabetes mellitus(T2DM), to identify the potential molecular specific markers of early T2DM. Methods: The blood samples of 4 type 2 diabetic patients and 4 normal control subjects were collected for microarray analysis. Then six candidate markers of LncRNAs screened from the differential expression profile were tested by qRT-PCR among the subjects (80 cases in the T2DM group and 50 cases in the control group). The possibility of these LncRNAs as molecular diagnostic markers was analyzed, and finally two of them were carried out by receiver operating characteristic (ROC) analysis. Results: Compared with control subjects, there were differentially expressed 133 LncRNAs in type 2 diabetic patients, among which 5 were up-regulated with the maximum up-regulated fold 3.29 and 128 were down-regulated with the maximum down-regulated fold 8.99. Six down-regulated LncRNAs were selected for validation and revealed a similar result to that of microarray.The expressions of two LncRNAs(NONHSAT160746 and NONHSAT140069) in peripheral blood of diabetic patients were significantly lower than those of control subjects (P<0.01). The areas under the ROC curve of the two LncRNAs were 0.734 and 0.703, respectively(P<0.01). Conclusion LncRNAs NONHSAT160746 and LncRNAs NONHSAT140069 are the potential molecular specific markers for the diagnosis of diabetes mellitus.