An integrated microfluidic system for automatic and self-validated analysis of cervical extracellular vesicle markers PD-L1 and ERBB3

Abstract

The early and precise diagnosis of gynecological malignancies, such as cervical cancer, is critical for improving patient treatments. Extracellular vesicles (EVs), such as exosomes, which carry molecular signals from their parental cells, offer a promising method for non-invasive liquid biopsy, however, conventional detection methods are often complex, high in reagent consumption, and susceptible to environmental fluctuations. To address this, we present an integrated, self-validated microfluidic system for the rapid, on-chip isolation and multiplexed identification of the gynecological EV markers PD-L1 and ERBB3. The chip achieved simultaneous on-chip processing of test and positive samples for parallel analysis within 1 h, enabling synchronous detection under the same conditions and thereby significantly enhancing the reliability of the assay. Additionally, a deep learning YOLOv8-based self-validated detection strategy facilitates automated and precise fluorescence identification. Validation with four cell lines (SiHa, C33A, HeLa, and H8) revealed remarkable EV protein signatures, achieving a limit of detection (LOD) of 15.56 particles/μL. This platform provides an integrated tool for sensitive and precise EV marker analysis, holding prospective potential for the early screening and personalized therapy guidance of gynecological tumor detection.