Angiogenesis and Artificial Intelligence, a long-overdue match
Since the invention of microscopy, researchers have solely relied on their visual and cognitive abilities to characterize complex microstructures found in biology. And rightfully so, as no feasible alternatives were available. With the emergence of computer-aided analysis, machine learning methods became more popular. However, employing this technology was cumbersome and the performance of the applications was still heavily influenced by the engineer’s ability to adequately instruct the computer to extract relevant information from the images. With the availability of more powerful processing hardware (GPUs), it was eventually possible to hand over feature extraction to more objective investigators – the computers.
Most recent advances in Artificial Intelligence (AI) rely on Deep Learning, a methodology used to build Neural Network models from raw image data. In other words, these algorithms are capable of learning visual recognition tasks automatically and independently from engineers’ experience. You may ask yourselves at this point: “How can I trust those applications to only learn relevant features and subsequently perform a reliable analysis?”. To master this, it is absolutely crucial to train the application on an evenly distributed, high-quality image dataset of your experiments. Even more importantly, you need to evaluate the performance on a test dataset that was not part of the training process. Because in the end, we do not want the model to learn “by heart”, but to generalize to unseen data.
Due to their nature, Deep Learning applications are a black-box and the exact process of how a model yields a certain result is often not evident. This may leave some researchers worried as an AI cannot easily articulate how it reached a certain result. However, when using Deep Learning for image analysis the issue of trust can easily be overcome as sophisticated applications also provide researchers with an interpretable visualization of the measured areas and not only the quantitative data. This feature facilitates traceability and trust in the analysis and allows researchers a simple cross-check whether the results are plausible. Another strategy when implementing Deep Learning applications is to run the new applications on an existing data set and compare the results in a benchmark. But be warned, you may soon find the Deep Learning application to yield superior results as it is not limited by subjective human perception.
Endothelial cell biology has recently found its rightful place in the spotlight and can greatly benefit from automated Deep Learning applications in many different research settings. For instance, these applications could support the investigation of feasible methods to block the formation of newly created blood vessels in hope of suppressing tumor growth. In another context, researchers are seeking strategies to vascularize synthesized tissue to overcome limits of diffusion, which is crucial for adequate nutrient supply. Using the latest Deep Learning applications they may now accurately investigate the process of angiogenesis and swiftly characterize the morphological features of vascular structures quantitatively.
Imagine you could easily determine the cell number in a cell culture dish from an image and accurately assess confluency within seconds. Deep Learning applications could support you to distinguish between various differentiation states of cells and determine their ratios over the course of the culture. Alternatively, with a little help, you could also locate and quantify spatial-distribution of cells in various matrices. This is by far not a complete list and you probably already have inspiring ideas yourself how your research could benefit from such Deep Learning applications. If so, get in touch!
Naturally, researchers and science itself would profit most from an intuitive tool that allows them to build Deep-Learning applications themselves and without any prior knowledge in computer science. Ideally, such a tool would be easy-to-use, fast, accurate and secure all while performing robust and reliable analyses. But who are we kidding, right?
Enough with daydreaming. Now let’s have a look at what assays we already have developed and can offer you right away.
Vascular network characterization has never been this easy by using our revolutionary image analysis applications readily available on the IKOSA Platform. If your research focuses on in-vitro research of angiogenesis we can instantly provide you with solutions to accurately quantify vascular structures in 2D and 3D matrices.
By employing the IKOSA Network Formation Assay Application all relevant vascular structure information from your 2D network such as area, length, branching points and number of tubes is automatically extracted from the microscopy image.
Are you currently studying endothelial cell behaviour in 3D scaffolds e.g. in response to certain signals? If so, we alleviate your image analysis tasks by providing you with applications to accurately assess crucial sprouting parameters of endothelial cell spheroids. Now you can rapidly quantify length, number, area and circularity of sprouts simply by executing the IKOSA Spheroid Sprouting Assay Application.
Additionally, you can take the investigation of 3D vascular networks to a new level by choosing the IKOSA Fibrin Tube Formation Assay Application and gain insight into vital parameters such as number, area, length and branching points of tubes in addition to area and perimeter of formed loops in 3D.
If you are currently performing ex-vivo research in angiogenesis using chorioallantoic membranes (CAM) we can provide you with state-of-the-art solutions to assess neovascularization in onplants by choosing IKOSA CAM Grid Assay Application.
Lastly, we are very proud to introduce you to the latest addition of our angiogenesis portfolio: IKOSA CAM Assay Application. Employing CAM Assay you are now able to automatically investigate vasculature formed on chorioallantoic membranes and gain valuable and quantitative insights into the area, length, branching points and thickness of vessels.
Do you believe a key application for investigating angiogenesis is missing or do you have developed new assays in your lab where you could benefit from sophisticated automated image analysis? We promise that you will! Give us a call +43 680 156 7596 or drop us a brief message firstname.lastname@example.org.
We would like to thank the following project team for the opportunity to use CAM Assay images (the first image of the blog post) in this article:
Dr. Nassim Ghaffari Tabrizi-Wizsy (Otto Loewi Reserach Center, Immunology and Pathophysiology, Medical University of Graz) provided the expertise of working with the CAM Assay, Lorenz Faihs performed the experiments and imaging/analysis, as well as DI Dr. Peter Dungel and A.o.Univ.-Prof. Mag. DDr. Kurt Schicho (Ludwig Boltzmann Institute for Experimental and Clinical Traumatology and University Clinic for Cranio- Maxillofacial and Oral Surgery, Medical University of Vienna), who planned the project.