CG-RecNet: A Gated and Attention-Fused Deep Learning Framework for Label-Free and Classification of Neural Stem Cell Differentiation via Imaging Flow Cytometry

Qinzi LiFang LiuJunYu ZhouXuanJian ZouChenLin GaoJingZe Li^*

• Sichuan Agricultural University, Ya'an, China

Precise and longitudinal monitoring of Neural Stem Cell (NSC) differentiation is pivotal for advancing regenerative medicine. However, traditional identification methods rely on invasive immunochemical staining, which terminates cell viability and precludes real-time analysis. To address these limitations, we propose CG-RecNet, a specialized deep learning framework designed for the accurate, label-free classification of NSC differentiation lineages—specifically neurons, astrocytes, and oligodendrocytes—directly from brightfield imaging flow cytometry (IFC) data. Unlike traditional microscopy, this approach leverages high-throughput single-cell imagery, focusing on fine-grained phenotypic classification. The architecture is explicitly designed to address the low contrast and visual ambiguity inherent in label-free imaging. It integrates a LinAngular Cross-Channel Attention (LinAngular-XCA) Fusion Module to capture global, long-range morphological dependencies (such as neurite extensions), and a Gated Convolutional Neural Network (GatedCNN) Block to selectively suppress background noise and refine local feature extraction. Validation on the internal test set from rat embryonic NSCs indicates that CG-RecNet achieves an incremental accuracy improvement of 1.82% compared to established baselines and state-of-the-art architectures, including DenseNet and Vision Transformers. The framework achieves an overall accuracy of 96.40% and a macro-average Area Under the Curve (AUC) of 0.9979. Notably, the model achieved high precision in identifying the minority oligodendrocyte lineage through feature refinement, offering an effective approach for handling the natural class imbalance of NSC datasets without synthetic oversampling. Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) analysis indicates that the model's attention aligns with biologically relevant morphological hallmarks, such as neurite outgrowth and soma texture. CG-RecNet provides a reliable, non-invasive, and qualitatively interpretable tool for neural stem cell research on the validated dataset.