RESUMO
Objective To establish a lumbar radiography image quality control model by using the deep learning algorithm and evaluate clinical images in real time and retrospectively based on the developed model.Methods The anteroposterior,lateral and oblique lumbar radiographs of 1389 patients collected between January 2018 to February 2021 at the The First Affiliated Hospital of Wenzhou Medical University were analyzed.The anatomical structures in the lumbar X-ray images were segmented using a full convolutional neural network based on U-Net,and the segmentation algorithm was utilized to establish an automatic evaluation model to detect substandard images.Dice similarity coefficient(DSC)was used to evaluate the performance of the model,and the lumbar radiography images were statistically evaluated after the application of the model.Results The accuracy of the model on the validation set was 0.971-0.990(0.98±0.10),the sensitivity was 0.714-0.933(0.86±0.13),and the specificity was 0.995-1.000(0.99±0.12).The quality control model had an excellent rate of 28.8%,an intermediate rate of 54.8%,and a failure rate of 16.4%for lumbar spine radiography in 2022.Conclusion The lumbar spine X-ray image quality control model based on artificial intelligence realizes accurate segmentation of lumbar spine anatomical structures and makes accurate evaluation of image quality,which is helpful to ensure the standardization of lumbar spine X-ray radiography operation by
RESUMO
Objective:To investigate the feasibility of tube potential of 80 kV combined with personalized contrast agent protocol in carotid artery CT angiography (CTA).Methods:A total of 136 consecutive patients undergoing neck CTA were prospectively enrolled in this study. The patients were randomly divided into Groups A, B, C and D. Tube potential of 100 kV and 15 s contrast agent injection protocol was used for Group A (53 cases) as conventional group, while tube voltage of 80 kV and 10 s contrast agent injection protocol was used for Groups B, C and D as experimental groups, with the contrast agent dosages of 20, 25 and 30 ml used according to the body weights of ≤50 kg(Group B, 20 cases), 50-70 kg (Group C, 38 cases), and 70-90 kg (Group D, 25 cases), respectively. The subjective and objective evaluation results of image quality and the effective doses were compared among the four groups.Results:The effective doses in Groups B, C and D were 1.54±0.91, 1.89±1.08 and 2.14±1.27 mSv, respectively, significantly lower than that in Group A [(5.66±0.56) mSv] ( F=169.34, P<0.05). The image quality of four groups met the requirements of clinical diagnosis. No significant differences were found in subjective evaluation and diagnostic efficacy of the four groups ( P>0.05). The CT number of carotid artery, signal-to-noise ratio and contrast-to-noise ratio of the neck region were significantly lower in Groups B, C and D compared with Group A ( F=14.9, 12.94, 14.43, P<0.05). The CT numbers of target carotid vessel were all higher than 250 HU. Conclusions:The scanning protocol of low tube potential (80 kV) combined with 10 s contrast agent injection protocol could not only reduce the doses of radiation and contrast agent, but also preserve the diagnosis effect. Thus, this scanning protocol was feasible and valuable in clinical application.