Efficient and versatile aberration correction through sensorless adaptive optics

Wavefront-sensorless adaptive optics methods are often used to correct phase aberrations in optical systems and thus to improve imaging quality. However, sensorless methods have an intrinsic disadvantage of requiring multiple images that can lead to non-desirable photo-bleaching. We have proposed a machine learning assisted aberration correction method which could correct aberrations consisting of not fewer than five Zernike modes with as few as two images. We showed that our method could be used in microscopes to provide instant aberration predictions when imaging biological samples of non-specific structures. We showed that compared to conventional function fitting sensorless adaptive optics methods, the new method corrected much faster with observable advantages. This novel method has a great potential to be used in any adaptive optics equipped microscopy for efficient sensorless aberration correction for biomedical microscopy.