Multicore fiber bundles for imaging and stimulating optogenetically modified neurons have been largely adopted in neurophotonics research. They allow for directed, single-cell stimulation and imaging of neuronal activity. An inherent limitation of these bundles is the presence and detection of the empty space between individual fibers, resulting in a loss of significant amounts of data, and reduced image quality due to pixilation effects. We propose a novel approach and algorithm to depixelation and image reconstruction from fiber bundles that utilizes multiple image frames collected during on-axis fiber bundle rotation. The approach involves first acquiring the Fourier transform of a stationary, unrotated image, followed by its rotated counterparts. The phase information from each image is then acquired, cross-correlated, and the angle of rotation determined from this correlation. Rotated images are then weighed and summed to generate a final reconstructed, depixelated image. Simulations were initially performed using Matlab demo images. Experimentation was done with a resolution chart, and thereafter with a cell culture. 488 nm and 561 nm continuous wave laser sources (Coherent, Inc.) were used for imaging GCaMP6s and C1V1-mCherry, respectively, in hippocampal neuronal cultures. The light sources were coupled to a multicore fiber bundle (Schott, 1534702) containing 4,200, 7.5 µm fibers. Cell cultures were prepared from 2 day old transgenic mice (GCaMP6s, Jackson Labs) transfected with C1V1(E122T/E162T)-TS-p2A-mCherry (Karl Deisseroth, Stanford). The results demonstrate this as an effective technique alongside fiber bundle imaging, serving as a useful and powerful tool for removing undesired artifacts associated with these fibers.
|