The MTAP model is a novel neural network architecture. While these related works also focus on using deep learning for brain tumour classification, the MTAP model distinguishes itself with a unique combination of model pruning, AdySYN for class imbalance removal and AvgTopK pooling for improved feature extraction. The resulting model achieves a state-of-the-art accuracy of 99.69%, exceeding the accuracies reported in studies on the Figshare dataset