Microfluidic eye sensors towards biomarker detection and biomedical applications

Abstract

Smart wearable biosensors have emerged as a promising technology for noninvasive and continuous health monitoring. This doctoral research presents the design, fabrication, and validation of a multifunctional microfluidic contact lens for rapid tear analysis in ocular diagnostics. The system integrated capillary burst valves (CBVs) and multifunctional biosensors within a soft, transparent PDMS substrate, enabling precise tear flow regulation and multi-parameter sensing in a compact and wearable format suitable for short-term (approximately 10 minutes) diagnostic use.

This research advanced the field through three main technical contributions. First, the microfluidic structure was systematically optimized to achieve reliable capillary-driven flow on a curved lens surface. Surface modification was applied to improve the wettability of PDMS and facilitate spontaneous liquid transport while maintaining optical transparency. Channel dimensions from 300 to 600 μm were comparatively evaluated, and 400 μm was identified as the most suitable size in terms of flow performance, structural integrity, and fabrication reproducibility. In addition, a wax and photocurable resin combined molding strategy was adopted for mold fabrication, which improved structural replication and enabled reliable formation of delicate microfeatures on soft lens materials.

Second, CBVs were incorporated to achieve controlled and sequential tear transport through interconnected microchambers. This design enabled the integration of tear volume, pH, and temperature sensing within a single contact lens platform while reducing cross-interference between adjacent sensing regions. Experimental results confirmed that sequential microfluidic control could be maintained on curved substrates, demonstrating the feasibility of combining capillary regulation and multifunctional sensing in a wearable ocular device.

Third, a machine learning–based analytical framework was established for quantitative interpretation of biosensor color responses under practical imaging conditions. Four models, including neural network, random forest, support vector machine, and K-nearest neighbors , were evaluated. Among them, random forest exhibited the most robust overall performance under small-sample and variable illumination conditions. The results indicated that machine learning could enhance the reliability of sensor data interpretation and support health-state prediction in portable settings.

Overall, this work demonstrates the feasibility of integrating microfluidics, multifunctional biosensing, and machine learning into a single wearable ocular platform. The integrated microfluidic contact lens achieved rapid and reliable tear-based diagnostics within minutes of wear, offering new possibilities for noninvasive ocular health monitoring. Although the current study was based on controlled experimental validation and independent image-based testing, further large-scale clinical evaluation remains necessary. Future work will focus on clinical translation, structural refinement, and adaptive analytical models for personalized diagnostic applications.

Date of Award	18 Jul 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Xu Sun (Supervisor), Nai Yeen Gavin Lai (Supervisor) & Qingfeng Wang (Supervisor)