Distortion and instability compensation with deep learning for rotational scanning endoscopic optical coherence tomography

• DOI: 10.1016/j.media.2022.102355
• Publication Date: 2022-01-22T16:23:44Z
• Authors: Guiqiu Liao; Oscar Caravaca-Mora; Benoit Rosa; Philippe Zanne; Diego Dall’Alba; Paolo Fiorini; Michel de Mathelin; Florent Nageotte; Michalina J. Gora
• Keywords: Distortion (music)

Optical Coherence Tomography (OCT) is increasingly used in endoluminal procedures since it provides high-speed and high resolution imaging. Distortion instability of images obtained with a proximal scanning endoscopic OCT system are significant due to the motor rotation irregularity, friction between rotating probe outer sheath synchronization issues. On-line compensation artefacts essential ensure image quality suitable for real-time assistance during diagnosis or minimally invasive treatment. In this paper, we propose new online correction method tackle both B-scan distortion, video stream shaking drift problem linked A-line level shifting. The proposed computational approach integrates Convolutional Neural Network (CNN) improve estimation azimuthal shifting each A-line. To suppress accumulative error integral also introduce another CNN branch estimate dynamic overall orientation angle. We train network semi-synthetic videos by intentionally adding rotational distortion into real images. results show that networks trained on data generalize stabilize videos, algorithm efficacy demonstrated ex vivo data, where strong artifacts successfully corrected.