This research presents a methodology and software for automating the analysis of optical coherence tomography (OCT) scans to facilitate the reconstruction of human blood vessels affected by atherosclerosis. Atherosclerotic plaque, a leading cause of cardiovascular diseases, requires precise diagnostics to guide effective treatment, but manual annotation of OCT scans, with over 300 frames per procedure, is time-consuming and is routing labor-intensive. Our approach combines traditional computer vision techniques and modern deep learning tools to streamline the preprocessing and annotation of OCT images. The proposed pipeline comprises four key steps: image calibration, anomaly detection and inpainting, probe shadow removal, and image annotation. Calibration ensures accurate mapping of pixel dimensions to metric units, critical for subsequent vessel analysis. For anomalies, such as embedded technical marks and doctors’ notations, we employ image thresholding, Hough line detection, and advanced inpainting using the LaMa model to reconstruct corrupted areas. Template matching is applied for the detection of probe shadows for further inpainting, reducing distractions in vessel segmentation. To accelerate annotation, we integrate Segment Anything Model (SAM), an advanced zero-shot segmentation model, enabling partial automation of vessel layer detection. This hybrid approach leverages OCT image similarities across patients, minimizing the need for extensive training datasets while maintaining high segmentation accuracy. The system is implemented in Python with a Gradio-based web interface, allowing seamless access and usability.