A Novel Spatio2-Frequency Blob Detection Algorithm for Enhancing Precision in Image Guided Surgery

Fluoroscopic images acquired during image-guided spine surgery (IGSS) procedures, serve as the input for pose calibration during guidance. A primary disadvantage inherent to fluoroscopic capture is a low-contrast image, making it difficult to accurately visualize and interpret the positional details of the fiducials that are required for calibration and distortion correction. This misinterpreted data directly affects the placement of pedicle screws. Even the non-detection of a small number of fiducials could result in a diminished distortion correction and increased calibration error, leading to a reduction in the precision of the IGSS procedures. In this paper, we propose a novel spatio
²
-frequency fiducial blob detection algorithm for varying contrast fluoroscopic images. This innovative fiducial blob detection filter has the following distinctive characteristics: (i) The algorithm quantifies the image quality from the C-Arm and adjusts the initialization parameters based on this image quality. (ii) Following the selection of initialization parameters, the fluoroscopic image is sent through a cascade of spatial and frequency filters for fiducial blob detection. The extent of cascading is determined by the contrast details of the image, as aggressive filtering will result in false object detection and noise amplification. (iii) The proposed filter can trim blob details for positional accuracy and remove outliers that could produce erroneous outputs. This inventive spatio
²
-frequency method was tested on standard in-house 350 fluoroscopic lumbar phantom images. The performance was compared on varying contrast and luminosity images to demonstrate the algorithm's efficacy in IGSS. The method successfully detects over 90% of fiducials in low-light fluoroscopic images, showcasing its versatility with the adaptive layered structure.