Revolutionizing blood diagnostics with real-time QPM technology

Blood tests are among the most common tools in medicine. Scientists are working to make blood cell imaging faster and more intuitive so that doctors can make fast and accurate diagnostic decisions. One promising technique, called quantitative phase microscopy (QPM), uses optical holography to measure the shape, thickness and size of individual cells without any dyes or contrast agents, which provides quantitative 3D information to assist diagnostic decisions. For diseases that cause changes in cell morphology, such as sickle cell disease (SCD), this method enables a high-throughput diagnostic tool that can be used at the point-of-care.

High-throughput QPM systems image flowing red blood cells (RBCs) at a high frame rate, acquiring images of over 100,000 cells in under 3 minutes. After reconstruction, researchers can conduct statistical analysis of large numbers of cells, which allows quantification of patient's SCD severity. However, to analyze phase images, the QPM data must be digitally reconstructed. The large amount of data collected by high-throughput QPM can take several hours to process on a regular CPU, whereas fast, real-time processing typically relies on expensive, high-performance GPUs, making it difficult to balance diagnosis time and cost in clinical applications.

A research team from Duke University developed a new real-time pipeline to reconstruct and analyze the high-throughput QPM data of RBCs at a rate of 1,200 cells per second. Their research is published in Biophotonics Discovery. This algorithm is implemented on a NVIDIA Jetson Orin Nano, an embedded GPU platform that only costs $249.

The processing pipeline is integrated with a high-throughput QPM system, which acquires and reconstructs imaging data of flowing RBC samples in real-time. It can automatically segment individual cell images, perform digital refocusing, and calculate each cell's morphological parameters such as volume and projection area, without the need for manual intervention during data collection. Researchers tested the system using polystyrene beads and healthy red blood cell samples. The real-time automated processing method reported highly accurate results, with an average error of less than 5 percent compared to traditional processing methods.

QPM has long held potential to provide detailed information about biological cells. Yet, the technique has yet to find widespread clinical use, often due to the cost or complexity in processing the imaging data. Here we have shown not only a high-throughput means for profiling thousands of cells at a time but also for rapidly processing and analyzing the information. This may be the missing step needed to bring QPM to the clinic."

Professor Adam Wax, leader of BIOS research group at Duke University

The authors suggest that the reported Jetson-based processing pipeline could benefit the development of a portable and low-cost QPM platform by balancing the diagnosis time and computational cost. Additionally, by applying an AI-assisted method, it is possible to conduct automatic blood screening in real-time and detect blood diseases such as SCD at an early stage.