Peak Detection Algorithm for Mass Spectrometry Integrating Weighted Continuous Wavelet Transform with Particle Swarm Optimization-Based Otsu

Abstract

Peak detection is an important step in mass spectrometry as accurately identifying characteristic peaks is key to data analysis. In order to address the issue of false peak detection, while simultaneously ensuring accurate detection of weak and overlapped peaks, this paper introduces an improved algorithm for mass spectrometry integrating weighted continuous wavelet transform with particle swarm optimization-based Otsu (WWTPO). The algorithm applies the weighted continuous wavelet transform (WCWT) to compress the frequency spectrum signal into a smaller scale range, which allows for the acquisition of more distinct and informative peak information. Moreover, the algorithm employs the particle swarm optimization (PSO) algorithm to iteratively evaluate the optimal image segmentation threshold, which addresses the challenge of inaccurate Otsu image segmentation. The method was applied to detect simulated peaks as well as matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) datasets. The performance evaluation was conducted using receiver operating characteristic (ROC) curves, F1 measure and F-scores. Through comparison with continuous wavelet transform (CWT) and genetic algorithm-based threshold segmentation (WSTGA), multi-scale peak detection (MSPD) and CWT and image segmentation (CWT-IS), the results demonstrate that WWTPO exhibits excellent performance in peak detection. The determination of 4-isopropyltoluene also demonstrates that WWTPO has excellent practical application. This method not only maintains a low false peak identification rate but also detects more weak peaks and overlapping peaks, further improving the accuracy and efficiency of peak detection in mass spectrometry.