An AI-enabled, low-cost microfluidic cytometry device for rapid cell analysis could make flow cytometry more affordable and accessible for low-resource and remote settings. Developed by a team at Rice University, the device demonstrates a level of accuracy comparable to conventional flow cytometers. Flow cytometry is a powerful technique for sorting and analyzing single cells, with applications in immunology, molecular and cancer biology, and virology. It is the gold standard lab test for clinical diagnosis and care and is used extensively in biomedical research. However, its use is limited to diagnostic labs and medical centers because it requires large, expensive equipment and complicated sample preparation and processing protocols. The AI-enabled, microfluidic cytometer developed by the Rice team is significantly cheaper and more compact than conventional cytometers, while requiring minimal sample preparation and instrumentation. To reduce the cost and size of the device, the Rice team used gravity-based slug flow in its design of the microfluidic cytometer. The device uses a microfluidic chip to drive the sample via gravity-based slug flow, eliminating the need for a pump and simplifying operation. Gravity-driven slug flow allows the sample to flow at a constant velocity through the microfluidic device, which is crucial for achieving accurate cell detection using an AI-based object detection technique. A slug flow-driven microfluidic chip developed by Rice University researchers. Courtesy of Doni Soward/Rice University. Gravity-driven slug flow is used primarily for transporting large volumes of liquids through industrial equipment. “To our knowledge, this is the first time gravity-driven slug flow has been employed for a biomedical application,” professor Peter Lillehoj said. The researchers used AI to enable the device to provide a quick, accurate count of CD4+ T cells — a marker of the body’s immune status — from unpurified blood samples. The device labels CD4+ T cells in the blood with anti-CD4 antibody-coated microbeads. An optical microscope and video camera record the sample flowing through the microfluidic chip. To accelerate image analysis and quantification, the researchers trained a convolutional neural network-based model to detect only the bead-labeled cells in the blood flow. An analysis of blood samples obtained from healthy donors showed that the microfluidic cytometer could quantify CD4+ T cells with accuracy comparable to traditional flow cytometers, with a less than 10% deviation between the two methods. The microfluidic cytometer is at least 4 times faster than traditional cytometers, with a rapid, 15-minute turnaround time. It is also less expensive and easier to operate. Due to its speed, portability, and ease of use, the microfluidic cytometer has the potential to be used for point-of-care applications for cell quantification. “Conventional flow cytometry is not practical for many resource-limited settings in the U.S. and around the globe,” Lillehoj said. “With our approach, this technique can be performed with ease for a fraction of the cost. We envision our innovative device will pave the way for many new point-of-care clinical and biomedical research applications.” The AI-enabled microfluidic cytometer can be readily modified to quantify other cell subpopulations by replacing the anti-CD4 antibody-coated beads with beads that are coated with antibodies targeting other proteins expressed on different cell types. The team envisions that the platform could also be modified for multiplexed cell quantification by using different colored antibody-coated beads. “Identifying and quantifying CD4+ T cells from unpurified blood samples is just one example of what one can achieve with this platform technology,” professor Kevin McHugh said. “This technology can be easily adapted to sort and analyze a variety of cell types from various biological samples by using beads labeled with different antibodies. “Based on the promising results we’ve obtained so far, we are very optimistic about this platform’s potential to transform disease diagnosis, prognosis, and the biomedical research landscape in the future.” The research was published in Microsystems and Nanoengineering (www.doi.org/10.1038/s41378-025-00881-y).