A Late Fusion Deep CNN Model for the Classification of Brain Tumors from Multi-Parametric MRI Images

Precise classification of brain tumors is crucial for legitimate clinical diagnosis, prognosis and treatment decisions. Multi-parametric magnetic resonance imaging (MRI) is an intriguing option for improving brain tumor classification accuracy. However, combining information from different MRI sequences poses a challenge, as each sequence provides unique and complementary information about the tumor. To overcome this challenge, this study proposed a late fusion CNN architecture that integrates features extracted from each MRI sequence at a later stage in the classification process. This approach facilitates the model to capture the unique features of each MRI sequence while also leveraging the supplemental data extracted from the other sequences. This study presents the implementation and evaluation of a cutting-edge deep learning-based late fusion multi-parametric brain tumor classification approach. The overall findings, alongside the feature maps, demonstrate how this strategy has the potential to enhance brain tumor classification accuracy and provide valuable insights for clinical decision-making. This investigation achieved 97% test accuracy.