Estimating Perturbations to Laser Position on Tissue for Lissajous Scanning in Endomicroscopy

This paper proposes an algorithm to reduce motion artifacts in Lissajous-scanned images taken from a custom-built endomicroscope based on microelectromechanical system (MEMS) laser scanning. The proposed algorithm uses an extended Kalman Filter to estimate uncertain perturbation to the position of the scanner relative to tissue during the Lissajous scan. The estimated states, which describe the estimated position perturbation to the scanner, are used to update the inferred laser trajectory during image reconstruction. To aid in estimation, the Jacobian matrix of a measurement function from other nearby points in the Lissajous pattern is computed using a trend of previous pixel values. For validation, the proposed algorithm was tested on both artificially generated blur images and real tissue images taken with the prototype confocal fluorescence endomicroscope. The reconstructed images with the estimated states showed reduced motion artifacts compared to the original images. The algorithm can be used in other single-pixel imaging applications that rely on beam steering technology to estimate the scanner's real trajectory and reduce motion artifacts.