Research on predicting radiographic exposure time in imaging based on neural network prediction models.

Objective: To explore the anatomical and clinical factors that affect the radiographic exposure time in radial artery cerebral angiography and to establish a model.

Method: A total of 210 patients who underwent radial artery cerebral angiography at this center from September 2021 to May 2022 were selected, and their anatomical and clinical factors were analyzed to evaluate the correlation between these factors and the duration of radiographic exposure. A related neural network prediction model was established.

Results: (1) Hypertension (p = 0.014), the presence of a double subclavian innominate artery curve (p＜0.001) and a proximal left common carotid artery loop (p = 0.014) increased the fluoroscopy time. (2) Age (p = 0.002), diameter of the aortic arch (p < 0.001), width of the innominate artery (brachiocephalic artery) (p < 0.001), and angle between the innominate artery and aortic arch (p = 0.042) were positively correlated with the fluoroscopy time. (3) The length of the innominate artery was negatively correlated with the fluoroscopy time (p = 0.024). (4) The accuracy of the neural network model was 80.0 % with the training set and 76.9 % with the test set, indicating a relatively high prediction accuracy. The ROC curve analysis showed an area under the curve (AUC) of 0.821, indicating its good predictive performance.

Conclusion: Age, hypertension, double curvature of the subclavian and innominate arteries, proximal left common carotid artery loop, diameter of the aortic arch, width and length of the innominate artery, and angle between the innominate artery and aortic arch can affect the fluoroscopy time for cerebral angiography in TRA.