A Robot-Assisted Optical Coherence Tomography (OCT) System for Inspecting Inner Wall Defects in Nonspherical Lumens

Inspecting defects in nonspherical lumens is challenging due to occlusion and depth variation, hindering complete point cloud data reconstruction. This article introduces a robot-assisted optical coherence tomography (OCT) system for conducting reconstruction and inner wall defects detection tasks in nonspherical lumens, leveraging robotic arm’s flexibility. Unlike image sensors, the developed OCT probe collects data via rotating scanning, complicating hand–eye calibration due to the lack of calibration features and coordinate drift in rotating probe. To address this, a decoupled hand–eye calibration method was proposed, utilizing iterative sphere-center fitting to improve accuracy in limited features and address coordinate drift problem. Moreover, OCT-acquired point clouds often include noise, artifacts, and exhibit anisotropy, complicating the detection of subtle defects. To overcome this challenge, we proposed a geometry-aware detection method based on local normal estimation to inspect defects in nonspherical lumens. Experimental results demonstrated a 55.2% improvement in hand–eye calibration with the proposed method, achieving a sphere fitting error of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$12.4~{\mu }$ </tex-math></inline-formula>m compared to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$27.7~{\mu }$ </tex-math></inline-formula>m in the traditional method. The method showed an average accuracy of 87.9% in detecting micrometer-level lumen defects, outperforming the state-of-the-art (SOTA) anomaly detection method (3D-ST), which achieved an average accuracy of 62.2%.