Modal-based algorithms for sensorless adaptive optics multiphoton microscopy

Over the last years multiphoton microscopy has been a useful tool for imaging biological samples. The technique offers optical sectioning capabilities and reduced photo-damage. However aberrations induced by the sample limit the performance of the technique especially at deeper locations. Sensorless adaptive optics has been used to overpass this. Instead of wavefront measurement, the procedure sequentially optimizes a suitable image quality metric by modifying the aberration corrector element. Although hill-climbing modal algorithms are often used for this purpose, alternative approaches might also improve multiphoton imaging and provide accurate solutions. Here we report a number of robust algorithms to minimize specimen-induced aberrations, paying special attention to ocular tissues. We implemented a Liquid-Crystal-on-Silicon (LCoS) phase spatial light modulator into the illumination pathway of a custom multiphoton microscope to improve the quality of images acquired at different depth locations. Individual Zernike modes were generated by a LCoS in order to pre-compensate the (plane-by-plane) sample’s aberrations. Two-photon excitation fluorescence (TPEF) and second harmonic generation (SHG) images were acquired and optimized according to a pre-defined metric. This optimization was carried out using three local search algorithms: deterministic Hill Climbing (HC), Stochastic Parallel Gradient Descent (SPGD) and Simulated Annealing (SA). Results showed that, for a common metric, the three algorithms (despite they work very differently) provided comparable optimized images in both contrast and sharpness, although the required wavefront aberration differed significantly. However, for the same Zernikes mode amplitudes and step increments, the HC algorithm requires 5 times more images than SPGD and SA to reach optimal solution. When static samples are involved the algorithm running time is not relevant, although photobleaching and phototoxicity might be present. However, when wavefront sensorless techniques are used in multiphoton microscopy of living specimens, faster procedures area more suitable to minimize acquisition times.