Research on Verification Methods and Tools for Medical Image Quality Control Standards

In order to solve the verification problem of standards related to image quality control in the field of artificial intelligence medical device, the relevant standard object is selected, the method of standard verification is studied, the framework of verification test tool is proposed, and the corresponding verification test tool is developed. In this paper, the quality of medical image is extracted as the verification factor according to the standard object. The image segmentation model is selected as the test object. The basic data set were obtained, and processed in order to obtain the data set of different quality as the input of the model test. Different data sets were input for model testing. The results of evaluation indexes such as Dice coefficient, Jaccard coefficient, Hausdorff distance and ASD distance were analysed. The evaluation index results show that different quality data sets have an impact on the model evaluation results. Further, the verification conclusion can be drawn that the data quality requirements proposed in the standard are scientific, reasonable and feasible. The research results can provide a feasible operational framework for the verification of international standards in other key areas of digital infrastructure. The design and development method of the verification test tool can also provide reference and guidance for the subsequent research of similar requirements.