Deep-Learning-Decoded Spatial Electrochemiluminescence from Potentiostat-Free, Button-Battery-Driven Microneedle Arrays for Multiplexed microRNA Imaging

Abstract Electrochemiluminescence (ECL) bioassays are essential for rapid intracerebral hemorrhage (ICH) diagnosis, while conventional approaches require bulky instrumentation and complex readout electronics that hinder point-of-care (POC) use. To address this limitation, we developed a battery-driven catalysis-initiated ECL imaging platform for multiplexed detection of ICH-associated microRNAs. The device integrated a gold film-modified copper microneedle array (Au film/Cu MNA), polydimethylsiloxane (PDMS) dual-chamber microcells, and button-battery excitation. Heterogeneous catalysis at the Au film/Cu MNA generated reactive oxygen species (ROS) that amplified ECL and sustained luminescence propagation across 168 μm. A competitive hybridization assay using ferrocene-modified linker DNA (Fc-LDNA) on microneedle tips enabled simultaneous miRNA-126 and miRNA-21 detection with detection limits of 4.6 and 0.8 pM, respectively. A hybrid convolutional neural network (CNN) model decoded spatial emission patterns, extending the linear ranges to 1.28 ∼ 32 fM and 0.59 ∼ 32 fM, beyond those achieved by pixel averaging. Preliminary clinical validation (n = 35 ICH, n = 25 healthy controls) revealed an inverse miRNA expression correlation and dual-target diagnostic accuracy (AUC = 0.993), exceeding single-biomarker performance (miRNA-126: AUC = 0.945; miRNA-21: AUC = 0.964). This strategy advances ECL bioanalysis toward portable, multiplexed biomarker quantification in neurological and inflammatory diseases.