Image-based adaptive optics for two-photon microscopy

Abstract

We demonstrate wavefront sensorless aberration correction in a two-photon excited fluorescence microscope. Using analysis of the image-formation process, we have developed an optimized correction scheme permitting image-quality improvement with minimal additional exposure of the sample. We show that, as a result, our correction process induces little photobleaching and significantly improves the quality of images of biological samples. In particular, increased visibility of small structures is demonstrated. Finally, we illustrate the use of this technique on various fresh and fixed biological tissues.

Fig. 1

Experimental setup: λ/2 (λ/4), half (quarter) wave plates; PBS, polarizing beam splitter; M, flat mirrors. Black solid lines, excitation path; dotted lines, emission path; gray lines, DM characterization path. Dashed lines show the conjugated Fourier planes. (b) Experimentally determined aberration modes Y_i used for correction.

Fig. 2

(a) Fresh mouse tissue before (left) and after one (middle) and five (right) iterations of the correction algorithm. (b) Image intensity and sharpness of the same sample as a function of the number of iterations.

Fig. 3

(a) Visibility improvement of small structures after correction. (b) Corresponding image of a fresh mouse liver sample 100 μm below the tissue surface before correction (top), after correction (middle), and before correction with an acquisition time twice longer (bottom). Inset, image of the same sample used for correction (same scale and color scale). Inverted contrast.

Fig. 4

(Color online) Depth-resolved correction on a mouse embryo. The two images (left, before; right, after correction) share the same color scale. Image size is ≈720 ×640 pixels. Insets, correction phase applied. (b) (Media 1, low res; Media 2, high res) 3D reconstruction of a 30×40 ×40 μm³ region of the embryo. (c) Aberration amplitude in Zernike modes as a function of depth. The objective correction collar was set so as to optimize image quality at a depth of ≈50 μm in the sample.