Traditional particle analysis methods often provide only limited insights beyond pure particle size and counts. AI-powered flow imaging microscopy enables a deeper, more concrete understanding of particles. By leveraging machine learning, Coriolis Pharma’s exclusive partnership with the developers of ParticleSentryAI enhances drug formulation, ensures product quality, and accelerates decision-making — leading to safer, high-quality biopharmaceutical products.
Moving Beyond Traditional Particle Analysis with Flow Imaging Microscopy
The presence of subvisible particles in drug products can significantly impact their safety and efficacy. While it is crucial to reduce the number of particles, it is — especially during drug development — an equally important task to further characterize the particles to understand their influence on product stability and performance.
Historically, light obscuration has been the standard technique for analyzing subvisible particles, but it provides only limited data, primarily measuring particle count and particle size without offering insights into their composition or structure. This limitation makes it impossible to differentiate between particle types and assess their impact on a drug product or withdraw information to guide process development.
Coriolis Pharma has been at the forefront of advancing flow imaging microscopy as a more powerful and sample-efficient solution for comprehensive particle characterization. By capturing images of particles, this method enables a deeper understanding of their size and morphology, including aspect ratio and intensity distribution. This capability not only allows researchers to distinguish between different types of particles — such as protein particles and silicone oil droplets — but also helps to track how formulation and process changes influence particle formation. The ability to analyze particles in greater detail empowers biopharmaceutical developers to refine formulations early in development, ensuring product quality and safety from the outset.
Advancing Particle Characterization with Artificial Intelligence
As one of the first service providers in Europe to offer flow imaging microscopy for particle characterization, Coriolis Pharma has generated a deep understanding of particle morphologies and built an extensive database of particle images over the last decade.
Historically, humans were required to interpret these images and differentiate particle populations within a given sample. This process is very time consuming and lacks repeatability, especially between different operators. To circumvent this, we frequently generated classification filters based on the morphological parameters provided by the imaging systems to automatically differentiate particle populations. While we have applied this approach successfully in many projects, it is time consuming to develop and it is limited by the few morphological parameters provided by the analytical instruments. Also, a differentiation of more than two populations is mostly impossible.
The integration of artificial intelligence (AI) is currently revolutionizing this process. By leveraging ParticleSentryAI, an advanced software solution developed by SentrySciences, Coriolis Pharma applies machine learning and computational statistics to generate unique “fingerprints” for each particle type. This AI-powered system employs a convolutional neural network to analyze heterogeneous particle populations, detect outliers, and classify subvisible particles with unprecedented precision.
While AI-driven particle image analysis does not change the way we capture the images — this is still done by the same flow imaging microscopy instruments — it can identify subtle morphological differences that are imperceptible to the human eye or impossible to find in classical morphological data sets. Similarly to how AI systems have outperformed even experienced radiologists in finding tumors on lung scans, AI-driven particle image analysis allows for a deeper and more reliable classification of particles, helping drug developers pinpoint the source of particulate matter, track changes in formulations, and optimize processes to ensure higher-quality biologic drug products.
Understanding Particle Morphology for Better Drug Development
Every particle in a drug formulation has a unique morphology, and understanding the morphological fingerprint of a particle population is important for ensuring product safety and efficacy. Coriolis Pharma is harnessing the power of ParticleSentryAI through a dual approach.
Our first approach is the development of generally applicable AI models. These models are designed to answer questions relevant to many drug product developers. For instance, our Coriolis model N°1 was trained on our in-house database to reliably differentiate protein aggregates, free fatty acid particles (from degraded polysorbate), silicone oil droplets, and air bubbles in a sample. This is a critical differentiation for any drug product presented in prefilled syringes, and we have successfully applied it already to improve drug product formulations and production processes for our clients. Our Coriolis model N°2 is designed as a research tool for cell viability. In our tests it was able identify live and dead cells, differentiate between apoptotic and necrotic cell death, and separate cell debris. Unlike traditional techniques, this analysis requires no staining or special sample preparation, making it a faster, more reliable alternative to other analytical methods.
These off-the-shelf AI models aim to be readily applicable to existing data sets, providing rapid, high-throughput particle characterization at a significantly lower cost compared to re-measuring samples.
By combining AI with flow imaging microscopy, Coriolis Pharma is pushing the boundaries of traditional particle analysis. As more high-quality data sets become available, the system can be trained to solve increasingly complex characterization challenges. Looking ahead, Coriolis Pharma is working on additional AI models to enhance understanding of protein aggregation pathways, such as those driven by heat or shear stress, and to characterize particles in other modalities, such as viral vectors and vaccines. New broadly applicable models are expected to be released within the next year, further expanding the capabilities of AI-driven particle characterization.
Expanding AI Models to Product-Specific Particle Characterization
Our second approach is the generation of custom AI models to answer questions specific to the product or process of our clients. As the exclusive licensed service provider for ParticleSentryAI in Europe, Coriolis Pharma is uniquely positioned to offer this cutting-edge technology as a service. While companies may purchase the software, only Coriolis has the expertise, experience, and exclusive rights to provide ParticleSentryAI as a fully integrated analytical service, making it the ideal partner for biopharmaceutical companies seeking advanced particle characterization solutions tailored to their product.
We have already successfully generated client-specific models, for example to support comparability assessment of drug product batches pre- and post-site changes. Here, the AI model generates precise fingerprints of the subvisible particle population inside a batch, and we can monitor any changes to this fingerprint between batches enabling a deeper understanding on how changes in the production process alter the particle population.
The AI can also be trained to differentiate particles originating from different product specific degradation pathways and identify those in stability sample to guide product and process development.
Based on our success with Coriolis model N°2, we hope that AI-enhanced flow imaging microscopy can play a role in cell-based potency or infectivity assays. These assays often rely on morphological changes that can be difficult to assess using classical light microscopy or ultraviolet and fluorescent-based methods. With AI-driven models, it may become possible to track subtle cellular changes following treatment with different formulations — offering a faster, more precise, and cost-efficient alternative for biologics development.
We are looking forward to applying AI based particle analysis to more product specific challenges and to understand if the AI models can meet our expectations. There are some current limitations to already consider. To generate a new AI model, a well-defined training dataset of at least a few thousand particle images is needed. For a reliable fingerprint comparison, typically a few hundred images are sufficient. This, however, is more than one would expect in stable, late-stage products. Coriolis is working closely with the developers of ParticleSentryAI to improve the tool and to overcome current limitations.
Achieving GMP Validation for AI-Driven Particle Analysis
While Coriolis Pharma continues to expand its AI models and capabilities, a key long-term goal is the development of GMP-ready particle analysis using ParticleSentryAI. SentrySciences is actively working to create a version of the software that meets regulatory validation requirements, and Coriolis is then hoping to demonstrate the applicability of AI based particle characterization under GMP in collaboration with clients.
Coriolis has already validated classical morphological filters for particle differentiation using flow imaging microscopy under GMP conditions. For example, filters were validated to classify particles based on size, roundness and intensity, allowing manufacturers to differentiate silicone oil droplets from other particulate matter and to determine whether a product meets the predefined quality specifications. The goal is to extend this validation process to AI-powered models, enabling the software to reliably identify specific particle "fingerprints" and determine whether a batch meets the defined specifications.
Flow imaging microscopy without AI is already recognized in USP Chapter 1787, “Measurement of Subvisible Particulate Matter in Therapeutic Protein Injections,” as a recommended technique for particle analysis. Looking ahead, regulatory agencies may increasingly expect drug developers to provide more comprehensive particle characterization data — beyond the basic numeric outputs — to ensure consistency and reproducibility in pharmaceutical products. AI-powered flow imaging microscopy has the potential to meet these evolving regulatory expectations, offering a more detailed and accurate approach to particle analysis in drug development and manufacturing.
Enhancing Drug Safety with AI-Driven Particle Analysis
Analytical methods used in drug development, manufacturing, and product release are critical for ensuring identity, purity, safety, and potency. Traditional particle analysis using light obscuration provides only quantitative data, measuring the number of particles in a sample without distinguishing between different particle types. This limitation is significant, as two samples may contain the same number of particles, yet those particles could vary widely in composition, origin, and potential impact on drug stability and safety.
AI-powered particle characterization using flow imaging microscopy offers a more comprehensive approach. By leveraging all available information from the particle images — beyond the parameters delivered by the instrument — this method enables a deeper understanding of particle composition and how it compares across samples. This level of insight contributes to improved drug safety and quality by identifying and mitigating risks that conventional techniques may overlook.
Applying AI-based particle analysis early in development can also identify potential issues sooner, helping drug developers understand which particles are problematic and where they originate. This proactive approach supports better candidate selection, product and process optimization, and ultimately reduces costs while enhancing overall drug quality and patient safety.
Driving the Future of Particle Characterization
Coriolis Pharma is transforming particle analysis with AI-powered flow imaging microscopy, providing deeper insights, greater accuracy, and faster, cost-effective solutions compared to traditional methods. As the exclusive service provider for ParticleSentryAI in Europe, Coriolis offers both ready-to-use and custom AI models, helping drug developers optimize formulations, improve product safety, and streamline regulatory compliance. With plans to expand its AI-driven capabilities in 2025, Coriolis remains at the forefront of innovation in biopharmaceutical particle characterization.