A Secure IoT-Enabled Machine Learning Framework for Brain Tumor Classification and Prediction Using MR Image Data

Abstract

Brain tumor identification and classification have improved due to the quick development of medical imaging and machine learning technology. This paper presents two approaches to secure image transmission in the Internet of Things (IoT): a comprehensive approach for brain tumor prediction and classification using a strong IoT infrastructure with cutting-edge machine learning models and a security approach with the implementation of the AES-ECC hybrid model in the MQTT communication protocol for image data encryption and decryption. We make use of a heterogeneous dataset that we sourced from the Kaggle Dataset platform, which includes four different types of MRI scans of brain tumors from 2870 patients. Our proposed methodology starts with the safe acquisition and transfer of MRI images through an IoT protocol infrastructure to a cloud-based platform. CNN, DenseNet, ResNet and G-Net are some of the sophisticated machine learning models that are used to interpret and analyse these pictures. The computer is trained to identify photos of brain tumors into the appropriate groups using all above four models. According to the data, our suggested CNN model performs better than the others, obtaining an amazing 89% accuracy rate. Nonetheless, we want to achieve even greater improvement in forecast precision by utilising ensemble boosting methodologies. Boosting the CNN model with Ada-Boost, Gradient Boost, XG Boost and Cat Boost algorithms aims to maximize prediction performance. We find that the CNN algorithm combined with XG Boost outperforms all other ensemble methods with an amazing accuracy rate of 97%. This encouraging result highlights how combining cutting-edge machine learning algorithms with IoT infrastructure can lead to better brain tumor classification and prognosis. The creation of more precise and effective diagnostic instruments for the identification of brain tumors is one of our study's many implications, one that will ultimately improve patient outcomes and the healthcare industry.