Infant Brain Tumor Detection Using Deep Learning

The difficulties in early detection and the potentially fatal nature of brain tumors in infants make them a major medical concern. The goal of this research is to create and test deep learning models that can automatically categorize pictures of brain tumors in infants. With a comparative investigation of their performance, two different neural network designs, DenseNet and AlexNet, are used for this purpose. The data set utilized for this analysis consists of pictures of brain tumors in infants, broken down into four types: pituitary, glioma, meningioma, and non-tumor. For data processing, model creation, and assessment, the project makes use of Python-based libraries including NumPy, pandas, seaborn, OpenCV, and TensorFlow. We use the state-of-the-art DenseNet201 architecture for our deep learning model. To improve their generalizability and robustness, both models are preprocessed and data augmented. A variety of measures, including as accuracy, precision, recall, and F1-score, are used to evaluate the model's performance during training and validation on the dataset. The findings show that DenseNet201 is more successful than AlexNet in classifying newborn brain tumors, with a weighted F1-score of 0.9703 and an accuracy of 97.048 % on the validation set. Early detection and intervention have the potential to improve patient outcomes in instances of newborn brain tumors, and our project's accomplishment in creating efficient and accurate deep learning models for infant brain tumor categorization bears this promise. Improvements to the models and their use in clinical settings to assist doctors with diagnosis might be in the works for the future.