CT-based radial endobronchial ultrasound image simulation of peripheral pulmonary lesions using deep learning

<bold>Objectives:</bold> Existing navigational bronchoscopy systems provide no guidance for R-EBUS. To guide R-EBUS manipulation, we aimed to simulate R-EBUS images of PPLs from preoperative CT data using deep learning. <bold>Methods:</bold> Preoperative CT and intraoperative ultrasound data of PPLs in 250 patients who underwent R-EBUS-guided transbronchial lung biopsy were retrospectively collected. 2D CT sections perpendicular to the biopsy path were transformed into ultrasonic reflection and transmission images using an ultrasound propagation model to obtain initial simulated R-EBUS images. A cycle generative adversarial network was trained to improve the realism of initial simulated images. Similarity between real and simulated images was evaluated. <bold>Results:</bold> There were no statistically significant differences in the long axis, short axis and area of the lesions between real and simulated images. Based on the experts’ evaluation, a median similarity score of ≥4 on a 5-point scale was obtained for lesion size, shape, margin, internal echoes, and overall similarity. <bold>Conclusions:</bold> Simulated R-EBUS images generated by our method can closely mimic the corresponding real images, demonstrating the potential of our method to provide guidance for R-EBUS manipulation. <bold>Fig. 1.</bold> (A1-A3) Original chest CT. (B1-B3) 2D CT sections perpendicular to the biopsy path. (C1-C3) Simulated R-EBUS images. (D1-D3) Real R-EBUS images. <fig><object-id>erj;64/suppl_68/PA1625/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig>