Research On Motion Artifacts Correction In Intravascular Photoacoustic Imaging

Intravascular photoacoustic(IVPA)imaging is a newly developed interventional imaging technique by multi-physics coupling for the diagnosis of vascular disease,especially coronary artery diseases.It provides structural and functional imaging of coronary arteries and atherosclerotic plaques.In the in vivo imaging of coronary arteries,the heartbeat and respiration as well as the pulsatile blood flow in the lumen lead to motion artifacts in the acquired dynamic image sequence,which are presented as the misalignment of vascular cross sections between adjacent frames in the transversal view and the saw-tooth shaped vessel wall boundaries in the longitudinal view.Motion artifacts lead to the reduction in the visual effect of images and the accuracy of image post-processing.Furthermore,the significant information for the diagnosis may be lost,thus reducing the accuracy of quantitative analysis of vascular morphology,structure and tissue components and three-dimensional reconstruction of vessels.This thesis focuses on suppression of motion artifacts caused by cardiac motion in IVPA image sequence.The main work includes two aspects as follows.First,an offline gating method based on affinity propagation(AP)clustering is proposed to correct motion artifacts.The sequential frames of PA signals acquired by continuous pullback of the imaging catheter are classified into static frames and dynamic frames by AP clustering.The gated image sequence is reconstructed from the static frames of signals,representing the optical absorption distribution on the vascular cross-sections.This method is capable of correcting motion artifacts before image reconstruction,avoiding post-processing of images.Second,the motion artifact correction based on deep learning is exploited by constructing a convolutional neural network called MAC-Net.The network is trained on computer-simulated dataset and then is used to optimize the dynamic frames of images.The motion artifacts are directly corrected without discarding any frames,ensuring the integrity of the pullback sequence.Both methods do not need synchronous ECG signal acquisition or online gating.The implementation is fully automated without any interaction of the operator.The demonstration results of numerical simulation indicate that the AP-based gating method is superior to the image-based gating in correcting heartbeat related artifacts with a lower computational cost.The visualization of IVPA longitudinal cuts is significantly improved by using the AP-based gating method or the direct correction based on MAC-Net.In addition,MAC-Net can not only improve image quality,but also ensure data integrity.