We present an attractive model of surface-enhanced Raman scattering-based lateral flow immunoassay (SERS-LFIA) for the sensitive and accurate detection of (MP) infection in human serum. The SERS-LFIA strip uses Au@Ag nanoparticles (Au@Ag NPs) loaded with two layers of Raman dye 5,5'-dithiobis-(2-nitrobenzoic acid) (DTNB) as SERS tags. The advantages of the dual dye-loaded SERS tags (Au/DTNB@Ag/DTNB) are the high sensitivity and the bioconjugation flexibility of the detection antibody. As determined from our SERS-LFIA strip, human IgM was quantified by monitoring the SERS signal on the test line. The limit of detection for human IgM was 0.1 ng mL, which was 100 times more sensitive than that by using the colorimetric method. Our assay results for 20 MP-specific IgM positive serum specimens showed 100% accuracy and detection rate, whereas the parallel enzyme-linked immunosorbent assay only showed 85% detection rate. The SERS-LFIA strip also exhibited high specificity and potential clinical applications. Therefore, our SERS-based LFIA strip has strong potential for practical applications in the sensitive and rapid detection of MP.