Deconvolution for multimode fiber imaging: modeling of spatially variant PSF

Wavefront-controlled light focusing through step-index multimode optical fibers (MMFs) enables minimally invasive endoscopic imaging of biological tissue. However, the spatially variant point spread function (PSF) of MMF imaging systems poses challenges for traditional deconvolution algorithms, which assume a uniform PSF.

To address this, we developed svmPSF, an open-source Java-based framework for ImageJ that models the spatially variant PSF as a series of spatially invariant PSFs. Using principal component analysis (PCA) on a series of point response measurements, svmPSF generates a precise PSF model. By integrating svmPSF with a modified Richardson-Lucy deconvolution algorithm, we successfully deblurred fluorescence images of beads and live neurons acquired via MMF, significantly enhancing image clarity and expanding the field of view (FOV).

This breakthrough in fiber-based endoscopic imaging improves high-resolution fluorescence microscopy, neural imaging, and biomedical diagnostics, paving the way for advanced minimally invasive optical imaging techniques.

Keywords: multimode fiber imaging, wavefront shaping, PSF deconvolution, spatially variant PSF, ImageJ, fluorescence microscopy, neural imaging, endoscopic imaging, minimally invasive optics, biomedical imaging.