A Chromatographic Peak Detection Algorithm Combining Dynamic Thresholding with Secondary Verification

Efficient data processing in gas chromatography-mass spectrometry (GC-MS) – particularly the automated and precise detection of chromatographic peaks – remains a critical bottleneck for the real-time, in-situ quantification of complex mixtures. Tailored to the unique operational parameters of portable GC-MS systems, this study introduces an automated peak-finding algorithm driven by dynamic thresholding and a secondary verification strategy. Initially, a nine-point moving average filter is applied to smooth the raw signal, effectively mitigating electronic noise while preserving crucial peak morphologies. Furthermore, first-derivative analysis and a window-averaged dynamic thresholding approach are implemented to adaptively isolate putative peak onset and offset boundaries. By integrating a contiguous valid-point criterion with a secondary verification protocol based on three-point adjacent comparisons, the algorithm successfully differentiates between isolated and overlapping peaks. Simulation analyses reveal that at a typical signal-to-noise ratio (SNR) of 20, the root-mean-square error (RMSE) for peak onset localization is only 0.0853 s. Notably, overlapping peak evaluations demonstrate that under low-resolution conditions (Rs < 0.8) across variable peak-height ratios, the success rate for overlapping point recognition exceeds 90%, substantially outperforming traditional derivative methodologies. Empirical validation using Toxic Organics-15 (TO-15) standard gas mixtures, alongside comparative algorithmic assessments, further confirms that this approach requires minimal computational overhead, inherently aligns with the sampling frequencies of portable devices, and accurately determines the integration boundaries of volatile organic compound (VOC) components. Ultimately, this framework significantly advances the automated analytical precision and robustness of portable GC-MS deployments in complex field environments.