Related Experiment Video
Updated: May 7, 2026
07:48
A High-throughput Automated Platform for the Development of Manufacturing Cell Lines for Protein Therapeutics
Published on: September 22, 2011
23.8K
A Modular and Scalable FPGA Platform for Intelligent, High-Throughput Image-Activated Cell Sorting
Yan Ding1, Jiehua Zhou1,2, Ruiqi Xi1
1School of Integrated Circuits, Wuhan University, Wuhan, China.
Summary
This study presents a stable, low-latency image processing solution for image-activated cell sorting (IACS) using field-programmable gate arrays (FPGAs). The system achieves rapid AI inference for enhanced cell classification and sorting accuracy.
Area of Science:
- Biotechnology
- Artificial Intelligence
- Microfluidics
Background:
- Image-activated cell sorting (IACS) links cell morphology to physiology for high-throughput classification.
- Integrating artificial intelligence (AI) enhances morphological analysis but increases computational complexity, leading to latency issues in IACS.
- High latency and instability in AI-driven IACS analysis can result in longer processing times and inaccurate cell sorting.
Purpose of the Study:
- To develop a stable, low-latency image processing system for IACS using field-programmable gate arrays (FPGAs).
- To enhance real-time AI inference for cellular data acquisition and classification within IACS.
- To overcome the computational challenges and latency problems associated with AI in IACS.
Main Methods:
- Developed a two-stage FPGA processing system leveraging the high-level synthesis (HLS) framework for cellular data acquisition and AI inference.
- Deployed a customized U-Net model on an AMD-Xilinx accelerator card, incorporating hardware acceleration for activation functions and convolutions.
- Implemented hardware acceleration modules, including activation function approximation and pixel-level convolution acceleration, for optimized performance.
Main Results:
- Achieved a stable image segmentation latency of 3.06 milliseconds (ms) at a 272 MHz clock frequency.
- Delivered a high processing throughput of up to 16,601 frames per second (fps).
- Successfully separated deformed HeLa cells and distinguished various cell types (colorectal cells, red blood cells, HeLa cells) and microspheres based on morphological parameters.
Conclusions:
- The developed FPGA-based system provides a stable and low-latency image processing solution for IACS.
- This approach significantly improves the efficiency and accuracy of AI-driven cell sorting.
- The system demonstrates the potential for real-time, high-throughput cell analysis in biological and medical applications.
