Label-Free and Intelligent Cell Death Recognition Toward Lung Cancer Chemotherapy

The lack of high-throughput, label-free, and intelligent recognition models for assessing cell death hinders the broad application of cell death analysis in chemotherapy for lung cancer. We propose an intelligent quantitative detection technique for cell deaths. Using high-throughput quantitative phase imaging flow cytometry to capture numerous label-free images and employing convolutional neural networks (CNN) to characterize the heterogeneity and quantitative detection of cell death. We revealed the heterogeneity of cell death through morphology features and achieved interpretability analysis of the CNN using clustering. Finally, the classification reliability of the CNN was validated by extracting features from classified cells. This method, compared with biochemical methods, showed a correlation of 0.92 and 0.91 with autophagy detection (Pearson and Cosine Similarity), and an average error of 12.52% with apoptosis detection. Our approach has the potential to become a valuable tool for studying cell death mechanisms and offers a new perspective for cancer treatment.