Optimized CNN Model for Early Detection of Brain Tumor from MRI Data

Abstract – Tumors in the brain be the reason for a significant number of deaths globally and can be grouped into various types with differing degrees of severity. Unfortunately, the survival period of more than 5 years is common only for 12% of adults suffering from brain cancer. In response to this Problem, this Investigation proposes a hyperparameter tuned CNN (Convolutional Neural Network) model aimed at precisely Recognizing brain tumors from Brain scans using a CNN. With regard to batch size, number of Levels, Learning speed, Activation layers, pooling, padding, and filter size, we have modified these parameters to improve feature extraction without increasing the model’s complexity. In order to test the efficacy of our system, we have trained our optimized CNN model on three publicly shared brain MRI-based datasets on Kaggle and received the following result: a typically of 76.14% accuracy across precision, recall, F1 score, and overall accuracy. Unlike the previously discussed methodologies, our model performed comparably to current state-of-the art methods and consistently showed improvements to performance and generalization. This provides an innovative approach that supports medical specialists in achieving brain tumor diagnosis with higher accuracy and efficiency. By improving the diagnostic process, the model enables faster, more dependable decisions and stands to have a favourable impact on patient care. Keywords – Brain Tumor, MRI, Deep Learning, CNN, Hyperparameter Tuning