Abstract
Colorectal cancer (CRC) remains a leading cause of cancer-related mortality, and detecting circulating tumor cells (CTCs) is crucial for early diagnosis and metastasis monitoring. Conventional staining-based cytology is costly, time-consuming, and often compromises sample integrity. In this study, we employed a combined digital holography (DH) and fluorescence imaging approach to develop a virtual staining framework for transforming quantitative phase imaging (QPI) data into interpretable pseudo-stained images. To the best of our knowledge, this is the first application of such a framework to colorectal cancer CTC detection. In our experiments, green fluorescent protein (GFP)-labeled HCT116 cells-generated through lentiviral transfection-were mixed with peripheral blood mononuclear cells (PBMCs) to create training datasets. The trained network achieved 99% classification accuracy and demonstrated strong generalization to unseen donors. This DH-fluorescence-based virtual staining method preserves cell integrity while enabling rapid, label-free, and low-cost liquid cytology diagnostics, highlighting its potential for non-invasive cancer detection and monitoring.