A study demonstrates that machine learning can significantly improve the precision of distinguishing between precancerous and cancerous human epithelial cervical cells using high-resolution atomic force microscopy (AFM) imaging of adhesion maps. The research, which utilized the random forest decision tree algorithm and K-fold cross-validation, showed an increase in the area under the curve (AUC), accuracy, sensitivity, and specificity compared to previous methods. This advancement is particularly important for clinical practice, as it could improve cervical cancer screening and reduce the number of unnecessary invasive biopsies.

The study explores the use of atomic force microscopy (AFM) combined with machine learning to identify geometrical features on the cell surface that correlate with cancer aggressiveness in human colorectal epithelial cells. The research demonstrates a novel approach to classify cells based on their surface characteristics, using Gaussian process regression and heatmaps to pinpoint regions indicative of high aggressiveness. The findings suggest that specific surface features, potentially microvilli and microridges, play a significant role in determining cell aggressiveness. The study highlights the potential of integrating physical and biochemical imaging techniques to enhance cancer diagnostics.