SBIR Phase I: Inline Monitoring of Particulate Matter and Sterility for Continuous Manufacturing

The broader/commercial impact of this Small Business Innovation Research Phase I project will be focused on a transformational approach to real-time quality assurance in injectable drug manufacturing by introducing a compact, low-cost and scalable sensor capable of detecting both particulate matter (PM) and microbial contaminants inline. Leveraging recent advances in imaging and machine learning, the system delivers automated, label-free particle analysis with high sensitivity and throughput. While standard quality tests are typically conducted offline and manually, this new technology fills a critical gap by enabling continuous, inline monitoring, supporting the industry’s shift from batch to continuous manufacturing. Real-time monitoring allows earlier identification of contaminants, reducing production downtime, minimizing waste, and improving product safety. Specific broader impacts of the research include strengthening U.S. competitiveness in smart manufacturing and quality assurance through advanced sensor integration and real-time analytics. The project is expected to cut recall and quality control costs by as much as 20%, potentially saving large manufacturers hundreds of millions annually. Its broader applicability spans food safety, environmental monitoring, and biodefense, offering scalable benefits for public health, ecological protection, and national security. The intellectual merit of this project lies in the development of a real-time, inline sensor system which integrates Digital Inline Holography (DIH) and deep learning for dual-function analysis of particulate matter (PM) and sterility in liquid production environments. Key innovations include high-throughput, label-free imaging; low false-positive detection of viable microbes; and robust PM classification with real-time visualization. DIH employs a low-power laser to illuminate particles in flow, generating interference patterns (holograms) captured by a high-resolution camera. These holograms are reconstructed into three-dimensional optical fields from which particle morphology, phase, and optical characteristics are extracted. A customized deep learning model, trained on those features using a diverse database of PM and biocontaminants, classifies contaminants at the single-particle level. The system also integrates a high-throughput preconcentration module to enhance detection sensitivity, achieving limits below 0.1 colony-forming units per milliliter (CFU/mL) at throughputs exceeding 1 mL/min. The modular, compact system design enables deployment at multiple stages of production. The deep learning model leverages transfer learning techniques, allowing efficient adaptation to new products and contamination types. This research will address key challenges in deploying optical sensors in industrial settings, including integration with fluid systems for robust, reliable operation under real-world conditions. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.