Deep learning (DL) has significantly advanced brain tumour diagnosis using Magnetic Resonance Imaging (MRI), yet most existing models suffer from high computational overhead and lack interpretability—key barriers to clinical deployment. We propose SpikeNet, a novel hybrid framework that integrates Convolutional Neural Networks (CNNs) with Spiking Neural Networks (SNNs) to address these challenges by combining spatial feature extraction with temporally sparse, biologically inspired computation. Evaluated on a brain MRI dataset, SpikeNet achieves an accuracy of 97.12%, precision of 97.91%, and recall of 97.65%, while reducing inference time by over 80% compared to Efficient.Net-B7, ResNet-50, and InceptionResNetV2. Moreover, SpikeNet produces high-fidelity saliency maps that better align with tumour regions than Grad-CAM and LIME, enhancing clinical relevance and trust. These results demonstrate SpikeNet’s potential as a fast, accurate, and interpretable AI system for real-time neuroimaging diagnostics.