EVAnalyzer2: Moving from in vitro to in vivo high content single vesicle imaging for quantitative serum and tissue pharmacokinetics of extracellular vesicles using C++ and artificial intelligence-based image analysis.

Introduction: With the rapid growth of the EV field, rigorous characterization and quantitation at the single vesicle level has become increasingly important to assure comparability and reliability of published data. To facilitate standardization, we recently developed ‘EVAnalyzer’, an ImageJ plugIn optimized for automated, quantitative image analysis from single vesicle imaging and cellular EV uptake data, which we published together with a robust protocol for routine EV applications. With > 3300 downloads, the program has become a tool of widespread use within the EV community.
Methods: Given this evident need in the field, we further progressed EVAnalyzer into EVAnalyzer2 that additionally enables tracking of EVs in huge, more complex in vitro, and in particular in vivo models. Usage of near infrared dyes, flexible image processing pipelines together with rigorous object quantification in these large histological sections enables the detection of EV signals over diverse (often spot/EV like) autofluorescent structures and thereby enables sensitive tissue and serum pharmacokinetics. In addition, advanced object recognition and segmentation is enabled by support of AI models, that are entirely based on open-source technology and can be easily trained and used without any deep AI knowledge. Moving into in vivo high content vesicle imaging, the increase in image sizes and the inherent growth in data volume make automated image and fast data processing inevitable and as these requirements hit the limits of the Java-based image analysis software Image J, we changed to the more efficient programming language C++.
Results: New developments for EVAnalyzer2 will be presented together with application examples for quantitative serum and tissue pharmacokinetics. Additionally, inspired by the rigor and standardization work done by ISEV, especially in the field of EV analysis by flow cytometry, we also evaluate and integrate calibration beads for single vesicle imaging, enabling comparisons of EV numbers and EV brightnesses between different instruments and imaging platforms.
Summary/Conclusion: EVAnalyzer2, an open-source high throughput image processing tool with a comprehensible user interphase allows the creation of highly flexible pipelines for standard image processing and thresholding, the inclusion of AI driven object detection and data processing options enabling single vesicle analysis of complex in vitro and in vivo images.