We present a deep learning approach for nuclear-level prediction of Ki-67 expression directly from H&E-stained breast cancer images, potentially eliminating the need for costly and time-consuming immunohistochemistry (IHC). Our two-phase pipeline integrates HoVerNet for precise nuclei segmentation with a specialized ResNet-style classifier (MitoNet) optimized for small patch classification. Our model, trained on 215,825 annotated nuclei, achieves 82.4% accuracy, with high sensitivity (recall: 0.925) in detecting proliferating cells. Comparative analysis demonstrates MitoNet’s superior generalization on external validation data (91.5% accuracy), while maintaining computational efficiency (2.1M parameters). Unlike previous approaches focusing on region-based predictions, our approach enables cell-level identification of Ki-67 positive nuclei, providing pathologists with granular proliferation assessment comparable to IHC staining.