DCNN and ResNet50-based Novel AI Framework for Automated Segmentation of Accurate Brain Tumor Diagnosis and Classification using Deep Convolutional Neural Networks

Abstract

The brain is the most important regulating system of the human body, and the abnormal growth of cells in the brain may result in brain tumors, which can cause serious health and life problems. Timely diagnosis and correct categorization are crucial to plan treatment and enhance prognosis in patients with these tumours. Modern diagnostic methods, including biopsies, imaging such as CT, PET, and MRI, are still very common, but they have drawbacks such as being invasive, subjective, and interpreted differently by different people. In response to these challenges, machine learning (ML) and deep learning (DL) systems for automated tumor detection, classification and segmentation in medical imaging have emerged as promising solutions. In this review, the authors emphasize recent developments of ML and DL techniques focusing on MRI-based tumor analysis. It compares traditional ML models like Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Random Forest, and Extreme Learning Machines, with the latest DL architectures, such as CNN, ResNet50, a 50-layer deep Convolutional Neural Network (CNN) architecture used for brain tumor detection by analyzing MRI scans to identify tumors with high accuracy, and transformer-based models. The work also covers frequently used datasets, evaluation metrics and segmentation methods, including simple thresholding and clustering, to more sophisticated deep learning frameworks. In addition, it highlights the need for further research and the need for increased attention to 3D modeling and attention mechanisms in enhancing diagnostic accuracy. The results show that deep learning models, especially DCNN-based, outperform traditional ML models in terms of accuracy, efficiency, and clinical reliability, making them suitable for brain tumor diagnosis.