Deep Learning-Based Detection, Classification, And Segmentation of Brain Tumours from MRI Images

The growth of abnormal cells in the brain is termed Brain tumor which is a serious problem because it leads to cause cancer. The usual methods to detect brain tumors are Magnetic Resonance Imaging(MRI) scans, Positron Emission Tomography (PET) scans, Diffusion Tensor Imaging (DTI) and Computed Tomography (CT) scans. Brain tumor segmentation from MRI scans is critical in clinical diagnosis and treatment planning. Accurate delineation of tumor regions aids in assessing disease progression and guiding therapeutic interventions. However, manual segmentation is time-consuming and prone to inter-observer variability. Hence, there is a growing interest in developing automated segmentation methods using deep learning techniques. Our methodology relies solely on CNNs for brain tumor segmentation. We begin by assembling a diverse dataset containing MRI images with various tumor characteristics. The used dataset is split into training and testing sets. We design a CNN architecture suitable for the task of tumour Classification, incorporating convolutional, pooling integrated with Relu and softmax as activation functions. The CNN is trained on the training set using stochastic gradient descent with backpropagation to balance the weight. The trained CNN is evaluated on the validation set, and hyperparameters are fine-tuned accordingly by using TensorFlow, NumPy, ipywidgets, sklearn Keras etc, in “Python� as it is an efficient programming language to perform fast work. Finally, we assess the performance of the CNN on the independent training and testing in the first training the Accuracy is 90.6% with 100 epochs and second training the accuracy is 99.4% with 400 epochs. this model is better than the state of the result obtained so far. This CNN-based model will help doctors detect brain tumours in MRI images accurately so that the speed of treatment will increase a lot.