Periodic Reporting for period 3 - DEEPVISION (Information-age microscopy for deep vision imaging of biological tissue)
Reporting period: 2019-03-01 to 2020-08-31
DEEP VISION takes a radically new approach to microscopy in order to lift this limitation: to use scattered light rather than straight rays for imaging. We use a technique called 'wavefront shaping' to make multiply scattered light converge to a sharp focus deep inside the turbid tissue. Once light can be focused at any point in the tissue, fluorescence microscopy can be performed by raster scanning the focus. This way, wavefront shaping has the potential to lift the limitations imposed by turbidity.
In the project we work on two critical problems that need to be overcome for this method to work. First of all, we developed and analyzed a method for focusing light deep inside scattering tissues completely non-invasively. Second, we discovered a property of scattering tissues that allows us to raster scan the focus efficiently. These ingredients, together, will ultimately result in a microscope that images at unprecedented depths inside scattering tissue.
DEEP VISION focuses on overcoming these two critical problems using wavefront shaping and other approaches. Since the research field of wavefront shaping is still young, it is not clear where lie the fundamental limits and possibilities of the techniques. In the first part of the project, we have worked on improving the understanding of wavefront shaping and light scattering in biological tissue, exactly to understand what are the limits and most promising future directions of wavefront shaping.
Our main results so far are:
- We found out how and when it is possible to form a focus inside scattering media non-invasively. With our simple statistical model, we can predict exactly under what conditions wavefront shaping will form a focus and when not.
- We found out exactly how far a focus can be scanned, depending on the turbidity of the sample and the design of the optical setup. Our work has united two types of wave correlations in scattering media into a single 'generalized memory effect'. Moreover, we developed a simple framework for analyzing, and predicting, and maximizing wave correlations in scattering media.
- We found that digital phase conjugation can be used to form a focus through scattering materials, even in the limit where each detector detects far less than a single photon of useful energy. Our finding is very good news for the community, we showed that the amount of required light is about ten thousand times lower than thought before.
- With the results above, we have mapped the possibilities and fundamental limitations of wavefront shaping. Our simple, yet accurate, analytical frameworks allow researchers in the field to easily calculate what are the fundamental limitations for a given system, and eventually develop a good intuition for what is possible and what is not.
- During the project, an extremely efficient numerical solver was developed for solving Maxwell's equations in large 3-D scattering media. Our method is 2-3 orders of magnitude faster than existing methods and up to 9 orders of magnitude more accurate. The solver is available as open source software, and it is increasingly being used by the optics community.
- The core aspect of DEEP VISION was to integrate imaging, modelling, and simulations in a closed loop system. This important milestone was reached in our demonstration of 'model-based wavefront shaping'. In this proof-of-concept experiment, we first mapped the 3-D structure of a light-diffusing surface using conventional microscopy (figure 1.1). Next, we fed this information into a computer simulation that calculated exactly how to focus light through the diffuser. (figure 1.2) This computed wave was generated experimentally, and used for high-resolution imaging (figure 1.3). Our results demonstrated an increase in imaging depth of a factor of 2 compared to state-of-the-art wavefront shaping (figure 2). These results introduce a new paradigm in wavefront shaping microscopy: computing a correction instead of iteratively measuring it.