Machine learning has developed rapidly in recent years.Using machine learning to predict the radiotherapy outcomes and complications can more accurately evaluate the patients' conditions and take appropriate treatment measures as soon as possible.The non-dose and dose related factors generated during radiotherapy are filtered and input into the algorithm model,then corresponding prediction result can be obtained.There are many algorithm models to predict survival rate,tumor control rate and radiotherapy complications,and the predicted result are more accurate now.However,the algorithm model also has various problems,and it needs constant exploration and improvement.

In recent years,the application of machine learning in the field of radiotherapy has been gradually increased along with the development of big data and artificial intelligence technology.Through the training of previous plans,machine learning can predict the results of plan quality and dose verification.It can also predict the multi-leaf collimator (MLC) positioning error and linear accelerator performance.In addition,machine learning can be applied in the quality assurance of intensity-modulated radiotherapy to improve the quality and efficiency of treatment plan and implementation,increase the benefits to the patients and reduce the risk.However,there are many problems,such as difficulty in the selection,extraction and calculation of characteristic value,requirement for large training sample size and insufficient prediction accuracy,which impede its clinical translation and application.In this article,research progress on the application of machine learning in the quality assurance of IMRT was reviewed.

Objective : To investigate the drug resistance and pathogenicity of six clinical isolates of Klebsiella pneu- moniae (Kp) ，and to provide a basis for prevention and treatment of Kp infection． Methods : The six strains from different hospitals were isolated ，cultured ，and identified by species-specific gene khe． Their whole genome se- quences (WGS) were obtained using next-generation sequencing technology (NGS) ．Based on the WGS，the cap- sular serotypes，sequence types (ST) and drug-resistance genes of six strains were identified．The capsular sero- type genes and virulence genes were validated or identified using PCR. Broth microdilution tests were conducted to validate their drug susceptibility，and mice were challenged with Kp aerosols by MicroSprayer aerosolizer to evaluate their pathogenicity. Results : The six strains were all serotype K2 but belonged to four ST types ( ST14 ，ST65， ST700，and ST86) ，and collectively carried six virulence genes and 23 drug-resistance genes．All the six strains were resistant to ampicillin，but only one strain was multidrug-resistant．Four strains exhibited high mucoid charac- teristics．Five strains could cause mortality in mice，which were preliminary identified as high virulence strains. Conclusion For the six Kp clinical isolates from different sources，only one strain named NY 13294 is both multi- drug-resistant and highly virulent，and other four highly virulent strains are resistant to one or two types of antibiot- ics.