Abstract: Bacillus atrophaeus (ATCC-9372) is an important strain of the Bacillus genus. The use of single particle mass spectrometry to distinguish unique biochemical markers of vegetative cells and spores of Bacillus atrophicus is important for understanding their biological properties. The main objective of this study is to distinguish vegetative cells and spores of Bacillus atrophaeus by analyzing the diameter and characteristic mass spectrometry ions of Bacillus atrophaeus by combined using of deep learning algorithms and classification model visualization methods. Firstly, the samples were prepared by collecting and centrifuging Bacillus atrophaeus that has been cultured for a certain period, and the spore samples of Bacillus atrophaeus were diluted. Then, single particle mass spectrometry was used to collect particle size and mass spectrometry data for the above two samples and to construct mass spectrometry datasets for the two objects. Following this, the particle sizes of the two samples were compared, and the datasets were divided. Based on the Matlab platform, a Convolutional Neural Network (CNN) classification model was trained to analyze the experimental results. Lastly, the typical ion characteristics of each were analyzed according to the average mass spectra, and the CNN classification process was visually analyzed using the Score-CAM algorithm. The differential ion characteristics between the vegetative cells and spores of Bacillus atrophaeus were extracted and analyzed. It was found that the particle size of vegetative cells is larger than that of spores, and the particle size of vegetative cells is essentially consistent at different sampling times. The CNN classification model achieves an accuracy of over 99% on both the test set and the validation set, indicating that the CNN model can fully learn and analyze the mass spectrometry characteristics. Their respective typical ion characteristics were analyzed by comparing the average mass spectra, which led to the introduction of their compositional differences, but not all typical ions could be accurately identified. Finally, a source analysis was performed on the ions with high scores in the Score-CAM results, and box plots demonstrated significant differences in the signal intensity of these high-scoring characteristic ions between the two states of Bacillus atrophaeus. Repeated experiments showed that the discovered high-scoring characteristic ions in the vegetative cells and spores of Bacillus atrophaeus have good stability and repeatability, suggesting their potential as species markers. This study performs an in-depth analysis of Bacillus atrophaeus in different states from a biochemical point of view, providing new insights into and methods for the processing and analysis of mass spectrometry data.