Brain Tumor Classification

Early classification and diagnosis of Brain Tumors are essential for providing the right treatment to a patient. It is crucial to get treatment as soon as possible because the survival rate for someone with an untreated brain tumor can range from as low as 3 months to as high as 5 years. In this project, we classified brain tumor images into 4 categories: glioma, meningioma, pituitary, and no tumor. With the use of baseline and deep learning models, the deep learning models demonstrated a significantly higher performance due to their ability to analyze images. The model with the highest accuracy was the MobileNet, a pre-trained transfer learning model trained on 5,608 images. This model yielded a validation accuracy of 98.24%. Using metrics including Kappa cohen score, precision, and recall, we validated the machine learning model's performance. We deployed the MobileNet model to a web app using Streamlit, where users submit MRI images and receive diagnoses of tumor class. We found that the model performed very well while utilizing the web app, indicating that it is safe to be used. However, since we only have 4 classes and there are over 150 total types of brain tumors, it could easily get a diagnosis wrong if it is not in one of these 4 classes.

Early classification and diagnosis of Brain Tumors is essential for providing the right treatment to a patient. It is crucial to get treatment as soon as possible because the survival rate for someone with an untreated brain tumor can range from as low as 3 months to as high as 5 years, which is why it is extremely important to get treatment as soon as possible. In this project, we used brain tumor MRI images to classify them into these 4 different categories: glioma, meningioma, pituitary, and no tumor. With the use of baseline and deep learning models, the deep learning models demonstrated a significantly higher performance due to their ability to analyze images. The model with the highest accuracy was the MobileNet model, a pre-trained transfer learning model trained on 5,608 images. This model yielded a validation accuracy of 98.24%. Using metrics including Kappa cohen score, precision, and recall, we validated the performance of the machine learning model. We deployed the MobileNet model to a web app using Streamlit, where users submit MRI images and receive diagnoses of tumor class. We found that the model performed very well while utilizing the web app, indicating that it is safe to be used. However, since we only have 4 classes and there are over 150 total types of brain tumors, it could easily get a diagnosis wrong if it is not in one of these 4 classes.

Explore More!

Source Code

Diabetic Retinopathy Evaluation: A ViT Approach

This study explores the use of a Vision Transformer (ViT) model to classify diabetic retinopathy severity from fundus images, demonstrating promising accuracy through transfer learning and data augmentation techniques. While results fell short of state-of-the-art benchmarks, the model shows potential for improving early diagnosis through AI-driven medical imaging.

Adi M.

Mentored by Abdullah Rashed Ahmed

Using Machine Learning Models to Predict Heart Disease and Attack with Non-Clinical and Non-Wearable Variables

This study uses data from the CDC’s Behavioral Risk Factor Surveillance System (BRFSS) to train machine learning models that predict heart disease risk based on non-clinical factors. The research demonstrates the potential for cost-effective, data-driven screening tools for early heart disease detection, though further improvements are needed for real-world application.

Samuel P.

Mentored by Rami Abi-Akl

Sleep Stage Classification: Investigating Deep Learning with High Dataset Imbalance Under Temporal Interval Change and Data Imbalance Techniques

This study investigates automated sleep stage classification using a CNN-GRU architecture, demonstrating that three-channel multimodal data and longer temporal contexts (60s and 90s) improve classification performance compared with two-channel and 30s approaches. The findings support ASSC as an efficient and accurate alternative to manual sleep scoring, with the best-performing model achieving a macro F1 score of 0.78 ± 0.02.

Ochir J.

Mentored by Joe Xiao

Brain Tumor Classification

Explore More!

Research Paper