We're excited to announce the release of VisualAITM software for protein biotherapeutics. An add-on for VisualSpreadsheet 6, VisualAI is a pre-trained, AI-driven image analysis module to recognize images of subvisible aggregates in biotherapeutics captured on any FlowCam 8100 or FlowCam LO instrument.
VisualAI is a powerful tool for quantifying protein aggregate and silicone oil droplet content in any protein-based drug product.
FlowCam with VisualAI provides a completely off-the-shelf yet powerful solution for differentiating subvisible particles in your protein formulations—no AI expertise, training data collection, image library preparation, or AI model training required. Out of the box, VisualAI achieves higher than 90% classification accuracy on particles larger than 3 µm in diameter. The software also includes features that can be used to identify anomalous particles such as calibration beads and air bubbles.
The graphical user interface of VisualAI (Click to enlarge): (Left) Sample image collage from a sample containing 50% protein aggregates and 50% silicone oil droplets. This sample was prepared as described in the White Paper. (Right) Collages in VisualAI’s graphical user interface showing the images identified by AI utilities that contain (top) protein aggregates, (middle) silicone oil droplets, and (bottom) other particles.
FlowCam with VisualAI provides the ability to streamline subvisible particle analysis in biotherapeutics by providing a uniform tool across different drug products and formulations. VisualAI is embedded in VisualSpreadsheet 6 and runs on any computer running VisualSpreadsheet – no need for powerful computing hardware.
In our new white paper “Robust AI Methods for Protein Biotherapeutics: VisualAI Software for Sample-agnostic Image Analysis With FlowCam” we present detailed performance results for VisualAI using images of different protein aggregates including monoclonal and polyclonal antibodies and silicone oil that were obtained with multiple FlowCam 8100 instruments.
We also demonstrate VisualAI’s effectiveness at measuring protein aggregate content in samples with known ratios of protein aggregates and silicone oil droplets.