ENHANCING BRAIN TUMOR DETECTION USING MRI WITH K-FOLD CROSS-VALIDATION

ABSTRACT A brain tumor, which happens when abnormal brain cells grow quickly, is a serious health risk for adults because it can lead to major organ problems and even be life-threatening. Manual segmentation of tumors from brain MRI is time-consuming and error-prone. Early detection allows doctors to get involved before extreme harm, bringing down permanent damage. In this paper, we will see the detection of tumors in Magnetic Resonance Imaging (MRI) where the Deep Learning model will be trained to detect whether the tumor is present. The dataset used in this study is sourced from Kaggle and contains two classes yes and no for brain tumor presence or absence. The tools and technologies used are Keras for developing the neural networks, scikit-learn used in data splitting and TensorFlow for managing data. Also, various libraries are used for image processing work, namely NumPy, OpenCV (cv2), etc. The model's performance is evaluated using an average of accuracy and precision as the primary metric. We aim to achieve accurate tumor detection by enhancing the model using K-fold cross-validation by converting the binary classification problem into categorical classification problem. We have trained the model using categorical cross-entropy for a binary classification problem, where the target variable has been converted into categorical format. The Saved model is used to predict the previously unseen data. Keywords— Deep Learning CNN, K-fold Crossvalidation, Magnetic Resonance Imaging (MRI), Brain Tumors.