A compact SpectroChip-based real-time reader for automated LFIA reaction-line localization and targeted spectral acquisition

This is the authors' abstract. We don't add key points for this paper.

Lateral flow immunoassays (LFIAs) provide rapid and low-cost point-of-care testing, but visual interpretation remains subjective, particularly for weak reaction lines. This paper presents a compact SpectroChip-based real-time reader for automatic LFIA reaction-line localization and targeted spectral measurement. The system integrates miniaturized spectral sensing, Raspberry Pi-based control, automatic region-of-interest selection, shutter-based exposure adjustment, dark- and white-reference correction, spectral smoothing, and one-dimensional derivative-based edge localization. Paired reaction-line boundaries are converted from video-frame indices into physical coordinates using frame-to-distance calibration and mechanical-error compensation. Validation with a multifeature calibration strip showed that all reconstructed line positions remained within independently measured reference intervals at both 60 and 90 frames/s. Under the fixed test configuration, the 90 frames/s condition produced lower coordinate variation than 60 frames/s, reducing the calibration-strip coordinate standard deviation to 0.043–0.055 mm and decreasing the average COVID-19 antibody LFIA C-line and T-line coordinate variations by approximately 66% and 75%, respectively. These frame-rate results are interpreted as a descriptive comparison under the present scan trajectory rather than as a general statistical superiority claim. Reliable C-line and T-line localization was achieved at the tested binding-antibody levels from 370 to 36 BAU/mL, whereas the 10-BAU/mL T-line response did not satisfy the predefined localization criteria. For the levels with successful T-line localization, the localized spectra were converted from corrected reflectance into an absorbance-like response and summarized using the AuNP-associated feature Δ A = A ( 530 ) − A ( 680 ) with a preliminary Hill-type response model; the 10 BAU/mL spectrum was treated only as a nominal-position low-response observation because its T line did not satisfy the automatic localization criteria. Across five complete measurement cycles, the reader generated localization and spectral results with an average turnaround of approximately 60 s, with measured times ranging from 56 to 80 s. The proposed reader therefore provides a compact and computationally efficient platform for automated LFIA localization, targeted spectral measurement, and preliminary characteristic-wavelength response analysis.