|
Endoscopic optical coherence tomography (OCT) is progressively used in endoluminal imaging because of its high scanning speed and near-cellular spatial resolution. Scanning of the endoscopic probe is implemented mechanically to achieve circumferential rotation and axial pullback. However, this scanning suffers from nonuniform rotational distortion (NURD) due to mechanical friction between the rotating probe and protecting sheath, irregular motor rotation, and so on. Correction of NURD is a prerequisite for endoscopic OCT imaging and its functional extensions, such as angiography and elastography. Previous work requires time-consuming feature tracking or cross-correlation calculations and thus sacrifices temporal resolution. In this work, we propose a cross-attention learning method for accelerating the NURD correction in endoscopic OCT. Our method is inspired by the recent success of the self-attention mechanism in natural language processing and computer vision. By leveraging its ability to model long-range dependencies, we can directly obtain the correlation between OCT A-lines at any distance, thus accelerating the NURD correction. We develop an end-to- end stacked cross-attention network and design three types of optimization constraints. We compare our method with two traditional feature-based methods and a CNN-based method, on two publicly-available endoscopic OCT datasets and a private dataset collected on our home-built endoscopic OCT system. Our method achieved a ~3 times speedup to real-time (26±3 fps), and superior correction performance.
|
