A rapid denoised contrast enhancement method digitally mimicking an adaptive illumination in submicron-resolution neuronal imaging.

Optical neuronal imaging often shows ultrafine structures, such as a nerve fiber, coexisting with ultrabright structures, such as a soma with a substantially higher fluorescence-protein concentration. Owing to experimental and environmental factors, a laser-scanning multiphoton optical microscope (MPM) often encounters a high-frequency background noise that might contaminate such weak-intensity ultrafine neuronal structures. A straightforward contrast enhancement often leads to the saturation of the brighter ones, and might further amplify the high-frequency background noise. We report a digital approach called rapid denoised contrast enhancement (DCE), which digitally mimics a hardware-based adaptive/controlled illumination technique by means of digitally optimizing the signal strengths and hence the visibility of such weak-intensity structures while mostly preventing the saturation of the brightest ones. With large field-of-view (FOV) two-photon excitation fluorescence (TPEF) neuronal imaging, we validate the effectiveness of DCE over state-of-the-art digital image processing algorithms. With compute-unified-device-architecture (CUDA)-acceleration, a real-time DCE is further enabled with a reduced time complexity.