Nanoparticle Diameter Estimation Using AI-based Models from Event Camera-based Optical Imaging

Description

Overview

This project aims to develop light-scattering microscopy methods for rapid, multi-attribute characterization of gene therapy particles at the single particle level. Emerging gene therapy treatments use gene delivery particles, such as viral vectors and non-viral carriers, to cure intractable diseases. But, because these small (20 - 200 nm) and complex nano-bioparticles are difficult to produce and accurately characterize, manufactured particles suffer significant heterogeneity issues, which hinder treatment efficacy and safety. To accelerate treatment development and improve efficacy and safety, we are developing light-scatter-based optical microscopes to provide physical measurements (e.g., mass, size, concentration, composition) of individual gene-delivery particles. 

The project approaches this challenge in two parts, nanoparticle diameter estimation and latency characterization of neuromorphic cameras. 

Nanoparticle diameter estimation 

Gene therapy delivery particles must satisfy rigorous safety and efficacy requirements; however, characterizing gene therapy delivery particles following manufacturing is a difficult measurement task. Quantifying features of these particles is challenging due to their small size (tens of nanometers to hundreds of nanometers) and variation in shape and composition. Single-particle tracking (SPT) via light-based microscopy provides a means to measure the size of nanoscale particles, which could be used to characterize gene therapy delivery particles, but this approach is limited by camera frame rate. The frame-based cameras, traditionally used for SPT, compromise the field of view (FOV) for frame speed. By contrast, event cameras can maintain a constant FOV while collecting data at high rates, improving particle trajectory statistics and reducing errors in size estimation. Our work addresses data analysis problems arising from the novel data format generated by event cameras in the context of nanoparticle tracking.

Early Findings 
This work addresses data analysis problems arising from the novel data format generated by event cameras in the context of nanoparticle tracking. Our approach includes investigations of baseline comparisons of frame and event cameras with reference materials, data analysis methods (filtering methods, tracking, target cost), and tradeoffs between imaging sensors at high sampling rates. We report results for a set of one-dimensional event filters to improve tracking fidelity, multiple lens magnifications, and an application-specific cost function while measuring gold particles with known diameters (51.4 nm +/- 4.1 nm) over a 24.4 second duration. To our knowledge, these results are the first application of an event camera toward quantitative nanoparticle characterization in the open literature.

Latency characterization of neuromorphic cameras

Pixel latency of event cameras is an important parameter for real-time applications. Pixel latency depends on the stimulus irradiance, the event camera configuration, and the number of active pixels. However, pixel latency is insufficiently described in current event camera specifications as a 'typical latency at a certain light level.' To develop a well-defined pixel latency measurement, NIST is working to define a hardware setup and a measurement procedure that (1) event camera vendors can reproduce to specify pixel latency and (2) event camera consumers can use to select event cameras for their real-time applications. Solving this problem will require (1) selecting accessible hardware for all vendors, (2) precisely specifying the hardware configuration and measurement procedure, and (3) defining the pixel latency metrics that are directly related to real-time applications. We describe such a solution and report pixel-latency measurements for three event cameras to demonstrate reproducibility and the practical value of this approach.

Early Findings
Our objectives in this project are to precisely define the characteristics of event cameras and the protocols for measuring them so that (1) event camera vendors can include them in their product specifications and (2) event camera consumers and integrators can determine whether the cameras meet their application needs. We completed the experiments with three event cameras.
 

Related Publications 

• Rémi V Chassagnol, Timothy J Blattner, Litorja Maritoni, Peter Bajcsy, “Pixel Latency Measurements of Event Cameras,” IEEE Transactions on Instrumentation and Measurement, October 21, 2025, https://ieeexplore.ieee.org/document/11210811; Print ISSN: 0018-9456 Online ISSN: 1557-9662 Digital Object Identifier: 10.1109/TIM.2025.3623767
• M. Cyrus Daugherty, Matthew DiSalvo, Jagat B. Budhathoki, Aaron Goldfein, Alexander Peterson, Thomas Germer, Ed Kwee, Gregory Cooksey, and Peter Bajcsy, “Nanoparticle Diameter Measurements With Event Camera Tracking,” Workshop on Event-based vision, held in conjunction with the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 11-15, 2025.

Created September 2, 2026, Updated September 18, 2026