Fast and Label-Free Accurate Identification of Immune Cell Classes by Short Integration Time Conventional Raman Spectroscopy Using Cascaded DAE-Resnet

Immunotherapy has achieved significant progress in the treatment of tumors and autoimmune diseases. The recognition and elimination of foreign pathogens and aberrant cells by the immune system are dependent on various immune cells, including T cells, B cells, and NK cells. Rapid, label-free analysis of these immune cells is essential for accurate disease diagnosis. However, traditional flow cytometry is limited by high costs and operational complexity. Raman spectroscopy is an optical method that operates without the need for labels, offering distinctive molecular fingerprint data for cells, but its application is constrained by the challenges of processing complex data with conventional analytical methods and the lengthy integration time required for data acquisition. To address these limitations, this study collected Raman spectral data with an integration time of only 1 second and employed a cascaded DAE-ResNet model to achieve rapid, label-free recognition of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T, B$</tex>, and NK cells with an accuracy of 96%. The results demonstrate that deep learning enables accurate immune cell classification under short integration times, offering a novel approach for high-throughput screening and clinical diagnosis with broad potential applications in biospectroscopy and disease diagnostics.