[4M1-GS-10] AI application: medicine

[4M1-GS-10-05] A Study on the Estimation Method of Peak Time of Test Contrast in CT Angiography Using Static Body Information

〇Toshihide Otsuki¹, Asahi Izumine¹, Kazuto Sakamoto², Homare Saisho², Hiroyoshi Yokoi², Toshitaka Yamakawa¹ (1. Univ. of Kumamoto, 2. Fukuoka Sanno Hospital)

Keywords:Contrast Effect, Machine Learning, Computed Tomography, Angiography

Computed tomography (CT) examinations measure the amount of X-rays transmitted through an object to obtain a cross-sectional image of it, and are generally divided into simple examinations that do not use contrast agent and contrast examinations that use contrast agent. In contrast examinations, the timing of imaging is an important factor in obtaining sufficient contrast effect and is determined by performing a test scan. In this study, we constructed a prediction model and evaluated its performance with the aim of reducing patient exposure dose by omitting the test scan and predicting the peak contrast arrival time using the patient's physical information. The mean absolute error (MAE) for the data used to validate the prediction model was 1.281. In addition, the relationship between the features that contributed significantly to the prediction and the peak arrival time was consistent with the results of previous studies.