OSA Incubator Explores the Future of Microscopy at the Royal Society, Edinburgh
Ivo Leite, University of Dundee, UK
The first international OSA Incubator - Future Microscopy: Merging Adaptive & Computational Imaging - took place today at the Royal Society of Edinburgh, Scotland, under the stern gaze of the nearby statue of James C. Maxwell. This meeting aims at discussing the future and challenges of emerging adaptive and computational imaging techniques in modern microscopy.
Traditional imaging systems make use of optical components which reconstruct an image of an object at a plane that is then sensed by an electronic detector and processed by computer to gain maximal information. However, a change in this paradigm is emerging in the field of microscopy. Image forming systems are being replaced by computational image reconstructions, and the integration of adaptive/reconfigurable optical elements into these systems allows for flexible and closed-loop operation.
Day One started with a welcome by Marcia Lesky, Deputy Senior Director of Business Development at OSA, followed by a short programme overview by the hosts Tomas Cizmar (Univ. Dundee, UK), Susan Cox (Kings College London, UK), and Brian Patton (Univ. Strathclyde, UK). The discussions started with a general introduction to the field of adaptive and computational microscopy by Martin Booth (Univ. Oxford, UK), Bernd Rieger (TU Delft, Netherlands), and Ernst Stelzer (Goethe Univ. Frankfurt, Germany). This included the imaging trade-offs in designing a microscopy system – imaged volume, time resolution, spatial resolution, and signal-to-noise ratio – and how adaptive optics can improve the latter two without compromising the others. The systems approach to adaptive and computational imaging (i.e. considering the system as a whole) can lead to non-conventional solutions not achievable by separate optimisation of each step in the traditional view of a linear processing chain. This means that the adaptive system is designed for testing a specific hypothesis, including specific sample preparation, but the image processing and results also feed back to adjust the hypothesis and modify the sample preparation. Closely related was the discussion of what defines the quality of an image: not much to argue here, usefulness always beats aesthetics in the science domain. The debate continued with the computational sources of aberrations, specifically in single-molecule localisation microscopy. Segmentation (finding regions of interest containing single-emitters) and localisation (fitting an adequate point spread function) are the main sources of image artefacts. Machine learning and maximum likelihood fitting were among the suggested solutions. This session ended with an overview of the main current challenges in microscopy: 3D imaging as a function of time of 3D specimens maintained under physiological (i.e. close to living) conditions, while addressing issues such as photo-bleaching, photo-toxicity and the natural energy levels of the sample. In this context, light-sheet based fluorescence microscopy (the Nature Method’s 2014 Method of the Year) is gaining increasing popularity.
The topic of spatial resolution dominated the rest of the morning. Susan Cox (King’s College London, UK), Allard Mosk (Utrecht Univ., Netherlands), Christian Soeller (Univ. Exeter, UK) and Omer Tzang (Univ. Colorado, US) led two panel discussions centred on how to best define it (especially in the context of localisation microscopy), and what are the obstacles to increase it. Perhaps surprisingly, the community has some difficulties in finding a general way to define a criterion for the resolution limit in computationally reconstructed images. Can the resolution performance of scanning (sampling) imaging systems be summarised in a single number? And can such a number (or set of numbers) be independent of the type of sample? Although no definite solution was reached, the participants seemed to agree that a set of standardised samples with varying sparsity – a standard “blob” resolution target was suggested – could be a good way to approach this problem. These and other related issues were intensively discussed – quite emotionally, at times – including the confusion of imaging resolution and localisation precision.
In the afternoon, Mark Neil (Imperial College London, UK) and Jonathan Taylor (Univ. Glasgow, UK) moderated a conversation on the challenges for deep tissue imaging. In particular, imaging in-depth in the ballistic regime was discussed, as well as the possibility of super-resolution imaging on highly dynamic samples. Still, some questions seemed to remain open. Is there a progressive route which the field can follow, or is a disruptive change needed? Is minimally-invasive endoscopy the best strategy for the near future?
Jacopo Bertolotti (Univ. Exeter, UK) and Michel Verhaegen (TU Delft, Netherlands) conducted a discussion on the influence of “prior information” on data analysis. When taking measurements researchers unavoidably make assumptions (the prior information): on the system, on the quality of components, on the theoretical models, sometimes on the sample itself. This does not mean we are necessarily doing something wrong, but not being aware of the assumptions we make and how they impact on our images (e.g. as artefacts) can render our data useless.
The first day of the OSA Incubator ended with a panel discussion led by representatives from three of the technology companies present at the meeting – M Squared (UK), Boston Micromachines (US), and Vialux (Germany). In addition to an overview of their most recent products, from specialised laser sources to deformable mirrors, optical modulators, and integrated adaptive-optics solutions, this gave the participants the opportunity to discuss the feasibility of the desired performance characteristics of some key technologies. While micro-electromechanical system (MEMS) based digital micromirror devices (DMDs) seem to gather increasing interest over liquid-crystal based spatial light modulators (SLMs), the community seems to crave a new modality of MEMS-based SLMs comprised of piston actuators, capable of providing fast continuous phase modulation. Also recurrently discussed was the possibility of integrating into commercial microscope systems a “knob” which could allow for the on-the-fly correction of aberrations (spherical + higher order spherical) using a reconfigurable component in the optical path. Perhaps next Christmas!
Stay tuned for Day Two of this Incubator Meeting!
Incubator hosts Tomas Cizmar (Univ. Dundee, UK), Susan Cox (Kings College London, UK), and Brian Patton (Univ. Strathclyde, UK) welcome the participants and lay out their goals for the meeting.
