A Convolutional Neural Network-Powered Dual-Mode Ratiometric Platform Integrating Ir-Doped Magnetic Metal−Organic Frameworks with Deep Learning for Intelligent Detection of Mycoplasma Pneumoniae

Abstract

Translating laboratory-grade quantification into field-deployable diagnostic tools remains a significant challenge in global pathogen surveillance. In this work, an artificial intelligence (AI)-integrated, dual-mode ratiometric biosensing platform is reported for the ultrasensitive identification of Mycoplasma pneumoniae (MP). The analytical mechanism is governed by a catalytic hairpin assembly (CHA)-mediated recognition cascade, which directs the specific spatial association of blue-emitting Ir/SiO2 nanospheres with red-emitting magnetic Ir/Fe3O4/ZIF-8 templates. This target-triggered nanoprobe assembly provides a self-calibrating ratiometric fluorescence signal for high-precision spectroscopic quantification while simultaneously generating a distinct macroscopic red-to-blue colorimetric shift. To eliminate subjective visual bias and the requirement for bulky instrumentation, a smartphone imaging setup is coupled with a convolutional neural network (CNN) to capture and decipher the complex optical transitions. By extracting specific chromatic features from digital images, the trained CNN directly converts visual phenomena into objective, concentration-dependent digital readouts. This proof-of-concept study demonstrates the feasibility of integrating ratiometric luminescent nanoprobes with deep-learning-assisted image analysis for intelligent pathogen detection. The spectroscopic ratiometric fluorescence mode is suitable for high-precision quantification, while the CNN-assisted smartphone imaging mode serves as a complementary semi-quantitative tool for field-oriented prescreening.