Label-free classification of single-cell lymphoma based on scattering imaging flow cytometry and machine learning

Label-free classification of single-cell lymphoma is essential for advancing cancer diagnostics. This study presents a method combining scattering imaging flow cytometry (SIFC) with machine learning to classify normal CD4+ T lymphocytes and T lymphoblastic leukemia (SUP-T1) cells based on distinct scattering patterns. Scattering data were captured using a microfluidic chip integrated with the laser-based SIFC system, and different lymphoma cells were classified based on the feature values of scattering images by means of machine learning. The results of five-fold cross-validation demonstrate that the model has excellent discriminatory power and reliability, with AUC values ranging from 0.945 to 0.964 (average: 0.957), validation set accuracies ranged from 86.3% to 92.0% (average: 89.0%), sensitivity, indicating correct identification of SUP-T1 cells, ranged from 89.0% to 93.0% (average: 91.6%), and specificity, indicating correct identification of CD4+ T lymphocytes, ranged from 83.0% to 91.5% (average: 86.4%). These findings indicate that SIFC, combined with machine learning, offers a promising approach for label-free, high-accuracy, and low-cost single-cell classification, with potential applications in cancer diagnostics.