Graduate student Swetha Lenkala, Master of Science in data science, will be defending her thesis titled: Deep Learning for the Diagnosis of Alzheimer’s Disease Using Ensemble Transfer Learning.
- Date: Monday, April 22
- Time: 1-2 p.m.
- Location: BARC 1122
Abstract:
Successful applications of deep learning emerged in many fields, including healthcare. Deep learning has the potential to provide accurate and efficient disease diagnosis, personalized treatment plans, and drug discovery. One of the medical conditions that started to benefit from deep learning is Alzheimer’s disease. Alzheimer’s is a progressive and degenerative brain disorder affecting memory, thinking, and behavior. However, even though there is no cure for Alzheimer’s, early diagnosis can help to improve longevity and quality of life. The biggest challenge in applying deep learning to medicine is scarce data resources for training models. Transfer learning is a technique that provides a solution to this challenge by utilizing pre-trained models from similar tasks and transferring the learning to the new task at hand. This reduces the amount of data and computation needed to achieve good results. Another technique utilized to increase the dataset size is to apply data augmentation techniques such as rotation and scaling. Deep learning approaches to detect Alzheimer’s utilizes various biomarkers, including brain MRI and PET images. Ensemble learning is an approach that merges the outcome of many machine learning models to have a final prediction with a better performance.
In this thesis, we work with MRI and datasets and apply various augmentation techniques to increase the dataset size. We utilized transfer learning techniques by fine tuning high-performing pre-trained models with augmented datasets. We applied ensemble learning techniques such as stacking and boosting to increase the performance of final predictions. To the best of our knowledge, no study utilizes data augmentation, transfer learning, and ensemble learning all together in one solution to provide better performance for Alzheimer’s diagnosis. We believe our thesis will contribute to the state-of-the-art machine learning approaches for Alzheimer’s disease diagnosis.
For more information, please contact Dr. Oguzhan Topsakal.