Smart Microscopy and Machine Learning for image acquisition
- Reference number
- RIF21-0043
- Project leader
- Camacho Dejay, Rafael
- Start and end dates
- 220901-271231
- Amount granted
- 15 000 000 SEK
- Administrative organization
- Göteborg University
- Research area
- Life Science Technology
Summary
A top challenge in quantitative biological imaging is to obtain large enough datasets to achieve statistical meaningful results. However, the need for human intervention in microscopy workflows limits the numbers of recorded events and the reproducibility of the results. Therefore, to tackle this challenge at the Centre for Cellular Imaging (CCI), we plan to develop smart microscopy workflows as a service for our users by connecting bio-image analysis, and computer-controlled microscopes to generate automated and adaptive imaging workflows. We will use on-the-fly image analysis to continuously monitor the sample, give real-time feedback to the imaging software, and then program the microscope to change its parameters during the experiment without human intervention. Further, we can train machine learning models to image samples at very low photon/electron budgets and predict biological events before they happen in ways that humans cannot, which can reduce phototoxicity in live-cell experiments and speedup image acquisition of large and complex samples. We believe that the challenges of smart microscopy are well suited to the capabilities of the CCI, which offers world-class access, service, and expertise to innovative imaging technologies since 2003. We will transform smart microscopy methods into services available to any researcher thanks to the high technical and scientific skill of our staff and the resources available within our infrastructure and international partners.
Popular science description
We all know the idiom “seeing is believing”, which has proven relevant not only on everyday life but also in scientific endeavours. Therefore, it shouldn't come as a surprise that microscopy has become a central technology in natural sciences, constantly providing important insights into cellular and molecular mechanisms throughout all branches of biological and medical research. Moreover, despite considerable technological advances, many microscopy experiments are still carried out in much the same way as they were decades ago, where samples are prepared and imaged one by one, and a scientist explores the sample until finding the region of interest before doing more detailed experiments. This need for human intervention in many cases hinders the acquisition of enough data to obtain statistically meaningful results, which is fundamental to sciences across all fields spanning both basic research in cell and developmental biology as well as translational research in tissue engineering, drug screening, and personalized medicine. Automation and artificial intelligence are revolutionising many sectors, from public transport to medicine. Implementing this vision in microscopy, referred to as “smart microscopy” requires a paradigm shift in current technology development and experimental design. It will no longer be the lone scientist sitting in front of a microscope that takes each image. It is the time to leave the path of designing microscopes for manual operation behind. It is not enough to merely make microscopes controllable via a computer, instead, image acquisition and image processing must be interconnected so that the microscope can adapt to the current view of the sample, make decisions based on artificial intelligence, and then acquire only what is essential to the scientific question and leaves the rest untouched. The Centre for Cellular Imaging (CCI) offers world-class access, service, and expertise to innovative imaging technologies since 2003. With our experience in advanced microscopy imaging, specimen preparation, data processing and analysis we support the broad needs of our users both in academia and industry. To tackle the challenge of smart microscopy at the CCI, we plan to develop microscopy workflows that use on-the-fly image analysis to continuously monitor the sample, give real-time feedback to the imaging software, and program the microscope to change its parameters during the experiment without human intervention.