Deep learning-based 3D leukocyte differentiation using label-free higher harmonic generation microscopy.

backgroundBoth in clinical practice and translational research, cell differentiation of leukocytes provides important diagnostic information and insights into pathophysiological mechanisms. The current gold-standard method for bronchoalveolar lavage fluid (BALF) analysis involves histochemical staining of cytospins, followed by manual morphological quantification. This approach however is labor-intensive, time-consuming, and highly operator-dependent, limiting its efficiency and throughput. This study proposes a deep learning framework for rapid, automated 3D leukocyte differentiation using label-free higher harmonic generation microscopy (HHGM).

methods3D leukocyte imaging was performed with label-free HHGM in a few minutes. Two deep learning models, ResNet 3D-50 and Vision Transformer (ViT) 3D, were trained, validated and tested for leucocyte differentiation on both BALF and blood fraction samples from 16 interstitial lung disease (ILDs) and 19 acute respiratory distress syndrome (ARDS) patients. Deep-learning model-prediction and cytospin analysis were performed by separate investigators. The results were compared using Bland-Altman analysis.

resultsThe proposed framework achieved accuracies above 86% for BALF and above 96% for blood samples under five-fold cross-validation. The approach shows close agreement with gold-standard cytological analyses, with mean differences of <5% across leukocyte subpopulations.

conclusionsBy integrating the label-free imaging capabilities of HHGM with deep learning, this study established a fast, accurate and high-throughput leukocyte differentiation in fresh BALF and blood samples. By significantly improving efficiency and reproducibility, this technology has the potential to transform clinical workflows and advance precision medicine.