Answers are grounded in this paper and cite their sources.
Distinguishing T-cell subsets is important for characterizing immune differentiation, but label-free classification is challenging because closely related subsets have overlapping biophysical properties. We combined holotomography with a three-dimensional convolutional neural network to classify primary human T cells from their refractive index (RI) tomograms. Fluorescence-activated cell sorting defined four reference populations: CD4+, naïve CD8+, and CD57− and CD57+ memory CD8+ T cells. The FishNet-inspired classifier achieved 97.04% accuracy in a cell-level held-out test of 270 cells. Dedicated binary classifiers achieved 97.70% accuracy for naïve versus CCR7− memory CD8+ cells and 94.44% for CD57− versus CD57+ memory CD8+ cells. Although the memory subsets overlapped in whole-cell descriptors, their volumetric RI patterns supported classification. These findings establish a basis for investigating intrinsic optical phenotypes associated with CD57-defined T-cell differentiation.
No takes yet. Share an insight, caveat, or question.
Ryu et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: