AI-powered image analysis to transform microscopy & biomedical imaging
Oeiras, 6th January 2025
Artificial intelligence is reshaping our daily lives, from how we navigate traffic to how we shop online. Now, a new AI-powered tool has the potential to significantly speed up the study, diagnosis and treatment of infection and diseases such as cancer, Alzheimer and other forms of dementia.
Traditionally, image processing tools have relied on static step-by-step instructions called algorithms. Now, ITQB NOVA researchers created a smarter system that uses machine learning to "learn" from data and make better decisions about which steps to take, depending on the task or the type of image it is working with. In simple terms, it is like moving from following a single recipe for every meal to having a smart chef who picks the best recipe for each specific ingredient and situation.
The research team has described this computational framework, called NanoPyx, in a new paper in Nature Methods. This AI-driven approach analyzes biomedical images up to hundreds of times faster than traditional methods, without losing accuracy. “Using this tool, we can reduce processing times from hours to minutes, or even seconds”, explains Bruno M. Saraiva, ITQB NOVA researcher and co-first author of the paper.
This advancement has wide-ranging implications for biomedical research and clinical applications. “One of the most exciting outcomes of our work is the potential for developing a new generation of smart microscopes”, remarks Ricardo Henriques, leader of the AI-driven Optical Biology lab at ITQB NOVA. “These advanced instruments would be capable of performing intricate analyses of the images they capture in real-time, allowing them to adapt dynamically to the sample.”
The researcher exemplifies how this applies to studying viral infection dynamics: “Typically, to observe how viruses enter and replicate inside host cells we need to capture time-lapse images over several hours and then spend hours or days processing them to achieve high resolution and extract quantitative data. However, with NanoPyx's "smart microscopy" capabilities, we can conduct super-resolution imaging and analyze viral particles as infection progresses”. This real-time feature ensures that the imaging system adjusts to ongoing events, enabling precise capturing of these interactions.
But NanoPyx’s potential can go beyond academic microscopy. By speeding up bioimage analysis, in the future, it could transform medical research, diagnostics, and personalized medicine. Faster analysis of biopsies and brain scans could lead to quicker, more accurate diagnoses and improved treatment strategies for cancer and neurological diseases, for example. The potential applications in drug discovery are equally promising: by accelerating the analysis of microscopy images to identify drug candidates, NanoPyx could bring new treatments to patients more quickly.
“NanoPyx represents a significant step forward in our ability to extract meaningful insights from complex biological data. As we look to the future, we are filled with excitement about the discoveries that this AI-powered tool might enable and the impact it could have on our understanding of life at the microscopic scale, potentially leading to breakthroughs in medical research and treatment strategies”, concludes Inês Cunha, co-first author from Stockholm University.
The study was developed in collaboration with the Bacterial Cell Biology and Intracellular Microbial Infection Biology lab at ITQB NOVA, led by Mariana Pinho and Pedro Matos Pereira respectively, with the Stockholm University (Sweden), the DFG Cluster of Excellence “Physics of Life”, TU Dresden and Turku Bioimaging, and the University of Turku and Åbo Akademi University (Finland).
Microscopy image resolution enhancement using methods in the NanoPyx engine, top left - image from microscope, bottom left - super-resolution computational prediction.
