Systems and methods are provided for obtaining raw mass spectrometry data from samples and generating an image representation of the mass spectrometry data. The image representation illustrates frequencies of occurrence of respective signals across the samples. A portion of the mass spectrometry data is obtained or extracted based on the frequencies. This portion is divided into segments. The segments correspond to different ranges or bins of retention times and mass-to-charge ratios. Veracities of signals within the segments are determined using a machine learning model. A subset of the segments that correspond to true signal is extracted. Mass-to-charge ratios and retention times corresponding to the subset of the segments are extracted.