Image-based adaptive optics for in vivo imaging in the hippocampus

Adaptive optics is a promising technique for the improvement of microscopy in tissues. A large palette of indirect and direct wavefront sensing methods has been proposed for in vivo imaging in experimental animal models. Application of most of these methods to complex samples suffers from either intrinsic and/or practical difficulties. Here we show a theoretically optimized wavefront correction method for inhomogeneously labeled biological samples. We demonstrate its performance at a depth of 200 μm in brain tissue within a sparsely labeled region such as the pyramidal cell layer of the hippocampus, with cells expressing GCamP6. This method is designed to be sample-independent thanks to an automatic axial locking on objects of interest through the use of an image-based metric that we designed. Using this method, we show an increase of in vivo imaging quality in the hippocampus.

Here, 4 correspond to defocus and 4 ( ) is an artificial defocus that allows to compute the PSF for the plane . The aberrations are represented by the vector = ( 1 , … , ).
The two-photon 3D PSF, also known as the two-photon focal volume, is ℎ 2 .

Definition of the metrics
For two-photon microscopy, the image intensity is given by the 3D-convolution of the 3D PSF ℎ 2 and , the fluorescence efficiency in the sample volume, called hereafter the "object".: The modal sensorless approach implies the optimization of an image-based metric computed on a transverse scan. One common metric is the total image intensity (Débarre 2009 S is the surface of the FOV. A pre-filtering can be applied to the image to enhance some particular structures. In our study, we considered a filter that enhances structures with a given characteristic size. The pre-filtered image for a transverse scan at 0 can be represented by the equation: ( , , 0 ) = ( , , 0 ) ⋆ 2 ( , ) For 10µm-size objects like neurons, F is an annular Gaussian centered at frequency 0.1µm -1 with a half width at 1/e of 0.1µm -1 .
Then, we can now define another metric, called hereafter pre-filtered image variance, as:

Figure S1: Changing the intensity related metric does not remove sample-dependency
Image quality metrics (no detection noise) as function of aberrations Z7 and Z11 (upper and lower graphics respectively) with a 10 bead in-focus and 12 out-of-focus (left and right respectively). The sample dependency can be observed for all metrics. Figure S2: Schematics of the two-photon laser scanning microscope