Brain Tumor Detection
Brain tumor detection and classification using MRI imaging is a crucial field of research in medical analysis. Every year, thousands of adults and children are hospitalized in the US due to brain tumors, making it vital for healthcare professionals to accurately diagnose and provide effective treatments. The combination of high-resolution 3D brain imaging and advanced machine learning techniques has greatly assisted medical professionals in this area. By utilizing MRIs and machine learning models, doctors can provide more precise diagnoses and personalized treatments based on the size, shape, and location of tumors. This integration of statistics and medicine is a rapidly expanding field of study, where image classification techniques in machine learning have shown promising results in aiding the treatment and management of malignant tumors.
In the spring of 2023, as a student in STOR 565 at UNC taught by Dr. Zhengwu Zhang, I gained knowledge about various machine learning algorithms. These included elastic net, support vector machines, K nearest neighbors, and neural networks.
During this time, I collaborated with four classmates (Kenza Slaoui, Simeon Kolev, Tanay Bali, and Jason Kennedy) on an exciting project involving the identification of brain tumors in MRI scans for our final project. To highlight our work, we created a team website that can be accessed here. Specific details of the project, including our final paper, can be found there.
In this project, my specific role was to develop models to accurately identify the presence of brain tumors in MRI scans. I conducted data preprocessing and implemented multiple algorithms to achieve accurate results. For my part of the project, I decided to use minimal preprocessing to ensure that my models would be applicable to the messiest MRI datasets. I tried out Support Vector Machines, a sequential neural network, and a random forest model. I utilized cross validation to find the best parameters for each model. Through the utilization of the random forest model, I was able to achieve an impressive test data accuracy of 96% for classifying the presence or absence of tumors in the test dataset.
Overall, this experience allowed me to apply my knowledge of machine learning algorithms in a practical setting and contribute to a meaningful project in the field of brain imaging analysis.