Non-invasive early diagnosis is of great significant in disease pathologic development and subsequent medical treatments, and microRNA (miRNA) detection has attracted critical attention in early cancer screening and diagnosis. However, it was still a challenge to report an accurate and sensitive method for the detection of miRNA during cancer development, especially in the presence of its analogs that produce intense background noise. Herein, we developed a surface-enhanced Raman scattering (SERS)–based lateral flow assay (LFA) biosensor, assisted with catalytic hairpin assembly (CHA) amplification strategy, for the dynamic monitoring of miR-106b and miR-196b, associated with laryngeal squamous cell carcinoma (LSCC). In the presence of target miRNAs, two hairpin DNAs could self-assemble into double-stranded DNA, exposing the biotin molecules modified on the surface of palladium (Pd)–gold (Au) core–shell nanorods (Pd-AuNRs). Then, the biotin molecules could be captured by the streptavidin (SA), which was fixed on the test lines (T1 line and T2 line) beforehand. The core–shell spatial structures and aggregation Pd-AuNRs generated abundant active “hot spots” on the T line, significantly amplifying the SERS signals. Using this strategy, the limits of detections were low to aM level, and the selectivity, reproducibility, and uniformity of the proposed SERS-LFA biosensor were satisfactory. Finally, this rapid analysis strategy was successfully applied to quantitatively detect the target miRNAs in clinical serum obtained from healthy subjects and patients with LSCC at different stages. The results were consistent with the quantitative real-time PCR (qRT-PCR). Thus, the CHA-assisted SERS-LFA biosensor would become a promising alternative tool for miRNAs detection, which showed a tremendous clinical application prospect in diagnosing LSCC.