Smartphone-Based Microbubble-Linked Immunosorbent Assay Powered by Classification–Regression Integrated Deep Learning for Portable Quantitative Biomarker Analysis

As the demand for portable, efficient, and precise diagnostic technologies continues to grow in modern healthcare, traditional ELISA systems are increasingly limited by expensive and proprietary equipment. Here, we developed CRiB DL -ELISA (Classification–Regression Integrated Deep Learning Microbubble-Based Enzyme-Linked Immunosorbent Assay), a smartphone-based platform for precise biomarker detection across 5 orders of magnitude. The biomarkers captured on the well plates are labeled with platinum nanoprobes, which generate distinctive bubble patterns in the presence of hydrogen peroxide. A user-friendly smartphone application integrates imaging, YOLO-based object detection, image preprocessing, deep learning inference, and result visualization to enable one-click analysis with real-time feedback. Concentration-dependent bubble patterns captured by the smartphone camera allow precise biomarker quantification without the need for additional optical hardware. The platform achieved a detection limit of 0.001 ng/mL for CRP and PCT, which was enhanced to ∼0.0001 ng/mL for NT-proBNP and cTnI via postprocessing. Quantification demonstrated high accuracy with R 2 = 0.9983 and 0.9979 for CRP and PCT, and R 2 = 0.9948 and 0.9942 for NT-proBNP and cTnI. Clinical validation showed 100.00% and 97.22% accuracy for CRP and NT-proBNP samples, respectively. Notably, NT-proBNP measurements showed strong concordance with commercial electrochemiluminescence platforms ( R 2 = 0.97), validating the clinical reliability. This CRiB DL -ELISA platform enables high-throughput biomarker quantification directly from microplates without specialized instrumentation, providing a cost-effective, robust, and portable solution for clinical diagnostics.