A false peak recognition method based on deep learning

Peak detection is a critical aspect of chemical spectrum data processing. However, existing peak detection methods are susceptible to producing false positive signals, which can complicate interpretation and lead to erroneous scientific findings. Manually detecting false peaks is a time-consuming process. Therefore this paper proposes a novel approach using a multi-scale convolution bi-directional LSTM attention depth network for identifying false peaks. This network leverages the feature extraction capabilities of the CNN network and the contextual connection abilities of LSTM. Our experimental results show that the proposed method achieves an accuracy rate of 89.67% in detecting false peaks in real LC-MS data, making it a powerful tool for automatic identification of false peaks.