Application of Artificial Intelligence in Single-Cell Mass Spectrometry Imaging

Single-cell mass spectrometry imaging (MSI) has transformed the fields of life sciences and medical research by facilitating the in-situ mapping of thousands of biomolecules. This advancement reveals the molecular distribution and heterogeneity that are essential for comprehending cellular function and disease mechanisms. As a label-free technique, MSI offers unparalleled insights into the spatial organization of metabolites, lipids, and proteins directly from biological samples. However, the performance of MSI has been significantly constrained by several limitations. A primary challenge lies in the inherent trade-off between spatial resolution and analytical sensitivity, achieving subcellular detail often compromises the detection of low-abundance molecules. Furthermore, factors such as variable ionization efficiency and ion suppression within the complex cellular microenvironment can obscure critical molecular signals. In addition to these analytical challenges, the technique generates vast, high-dimensional datasets. Each experiment produces spectra from hundreds of thousands of pixels, resulting in an immense volume of data that is exceptionally difficult to process and interpret using traditional statistical and bioinformatic methods. The rapid advancement of artificial intelligence (AI) is currently revolutionizing the field of MSI analysis. AI algorithms are particularly well-suited to tackle these complexities and are being seamlessly integrated throughout the analytical pipeline. Their advantages in image fusion, data dimensionality reduction, pattern recognition, and complex modeling have become indispensable. For example, AI-driven super-resolution and denoising techniques significantly enhance image quality beyond the optical limits of instruments. In terms of data analysis, AI excels at non-linear dimensionality reduction, effectively unraveling intricate datasets to reveal intuitive patterns and distinct cellular subtypes. Moreover, convolutional neural networks (CNNs) automate pattern recognition processes, enabling precise classification of tissue types and identification of disease biomarkers with exceptional accuracy. Additionally, AI facilitates the integration of MSI data with other omics fields, promoting a more comprehensive systems biology approach. In light of this transformation, this article systematically reviewed recent applications of AI in mass spectrometry, with a particular emphasis on single-cell MSI. The review was organized around the predominant ionization techniques that facilitate this analysis, including matrix-assisted laser desorption/ionization (MALDI), secondary ion mass spectrometry (SIMS), desorption electrospray ionization (DESI), and vacuum ultraviolet laser desorption/ionization (VUVDI). It examined how AI models are specifically designed to harness the unique advantages and address the limitations inherent to each method. Meanwhile, the technical progress, challenges, and future prospects were also discussed. Finally, future prospects for this potent synergy between artificial intelligence and spatial metabolomics were delineated. This collaboration was poised to unveil new frontiers in biomedical research.