AI Virtually Stains AFM Images, Revealing Cell Phenotype at Subcellular Detail

Atomic force microscopy (AFM) has recently emerged as a powerful tool for identifying the malignancy of cells with high precision. A recent study utilizing AFM Ringing mode imaging of human colorectal epithelial cells demonstrated the ability to distinguish cells with varying degrees of cancer aggressiveness through machine learning (ML) analysis. However, traditional ML methods analyze entire AFM images, lacking the ability to pinpoint specific cell surface features associated with increased cellular aggressiveness. To address this limitation, we propose a novel machine-learning approach capable of identifying discrete geometrical features on the cell surface that are indicative of highly aggressive cell classifications. By applying our ML algorithm to AFM Ringing mode images, we enable the virtual staining of cells, highlighting phenotypic differences with subcellular resolution. This targeted approach provides valuable insights into the morphological characteristics linked to cancer aggressiveness. The biological implications of the identified cell surface features are discussed, shedding light on potential mechanisms underlying aggressive cancer cell behavior. The application of this technology to other type of cancer cells is also presented. Our findings demonstrate the utility of combining AFM imaging with advanced machine learning techniques to enhance the characterization and understanding of cellular abnormalities at the subcellular level. This innovative approach holds promise for improving diagnostics and contributing to personalized medicine.