AI-Enhanced Microfluidic Chip Enables Label-Free Detection of Circulating Tumor Cells

Source: RadAISlice (September 29, 2026). AI-generated summary by biochip.com, published October 1, 2026. Not independently reviewed.

Researchers have combined a microfluidic chip with artificial intelligence to isolate and identify rare circulating tumor cells from blood without fluorescent labels. The system separates cells by their physical size, then uses ordinary bright-field microscope images and the YOLOv8 deep learning model to recognize likely tumor cells. In testing with artificial blood samples, it recovered MCF-7 breast cancer cells at a rate of 89.2% while removing 86.9% of white blood cells. The work points toward a simpler workflow for studying tumor cells that travel through the bloodstream, though it still needs validation using patient samples.

Sorting Rare Cells by Their Physical Properties

Circulating tumor cells, or CTCs, are cancer cells that have detached from a tumor and entered the bloodstream. They are difficult to find because a blood sample contains vast numbers of red and white blood cells, while tumor cells may be scarce.

The chip uses contraction-expansion inertial microfluidics to enrich the sample for CTCs. Think of it like guiding a mixed crowd through a corridor that repeatedly narrows and widens: different-sized travelers follow different paths. Inside the chip, fluid forces sort larger tumor cells away from many smaller blood cells without requiring chemical tags attached to the cells.

AI Reads Standard Microscope Images

After the chip concentrates the potential tumor cells, the system captures bright-field images, the standard images produced when visible light passes through a sample. This avoids fluorescent staining, a method that can require extra reagents and depends on markers found on a cell surface.

The researchers used YOLOv8, a deep learning model designed to locate and classify objects in images, to identify tumor cells in those microscope images. In their evaluation, the model reached 96% accuracy, with 94.6% precision and recall and 95.4% specificity. Precision measures how often cells identified as tumor cells were correct, while specificity reflects how well the system avoided incorrectly labeling non-tumor cells as tumor cells.

Label-Free Detection Could Simplify Workflows

Many CTC workflows use antibodies or fluorescent dyes to identify cells with particular biological markers. Those approaches can be useful, but they add preparation steps and may miss tumor cells that do not display the selected marker.

By relying on physical separation followed by image-based classification, this platform aims to detect cells without cell-surface marker labeling or fluorescent staining. The chip's measured MCF-7 recovery rate was 89.2% plus or minus 3.1%, while its white blood cell removal rate was 86.9% plus or minus 1.4%.

Why This Matters

A blood draw is far less invasive than collecting tissue from a tumor, making circulating tumor cells an important target for liquid biopsy research. A label-free microfluidic system could potentially make rare-cell enrichment more streamlined and preserve cells for subsequent analysis.

Still, the reported validation used artificial blood samples containing MCF-7 cells, a human breast cancer cell line. Testing the combined chip and AI workflow with real patient samples will determine how well it handles the biological variation, low cell counts, and visual complexity of clinical blood specimens.

The next studies will therefore need to move from controlled samples to patient-derived CTCs. If performance holds in that setting, the pairing of inertial microfluidics and bright-field image analysis could offer a practical route to label-free rare-cell detection.

Topics: Microfluidics · Cancer

Start here: Microfluidics explained