Automatic correction of the initial rotation angle error improves 3D reconstruction in endoscopic airway optical coherence tomography.

Abstract

Endoscopic airway optical coherence tomography (OCT) is an advanced imaging modality capable of capturing the internal anatomy and geometry of the airway. Due to fiber-optic catheter bending and friction, the rotation speed of the endoscopic probe is usually non-uniform: at each B-scan image, the initial rotation angle of the probe is easily misaligned with that of the previous slices. During the pullback operation, this initial rotation angle error (IRAE) will be accumulated and will result in distortion and deformation of the reconstructed 3D airway structure. Previous attempts to correct this error were mainly manual corrections, which are time-consuming and suffered from observer variation. In this paper, we present a method to correct the IRAE for anatomically improved visualization of the airway. Our method derived the rotation angular difference of adjacent B-scans by measuring their contour similarity and then tracks the IRAE by formulating its continuous drift as a graph-based problem. The algorithm was tested on a simulated airway contour dataset, and also on experimental datasets acquired by two different long range endoscopic airway OCT platforms. Effective and smooth compensation of the frame-by-frame initial angle difference was achieved. Our method has real-time capability and thus has the potential to improve clinical imaging efficiency.