Primary supervisor
Fuad NomanResearch area
Machine Learning and Deep LearningArtificial Intelligence (AI) is revolutionizing the field of Magnetic Resonance Imaging (MRI) by enabling faster, more accurate, and cost-effective image reconstruction. This project explores cutting-edge AI methodologies, focusing on combining data-driven approaches with physics-informed models to tackle challenges in MRI reconstruction. By integrating MRI acquisition physics directly into neural networks, we aim to improve the interpretability and robustness of reconstruction techniques. Another exciting direction involves self-supervised learning, which reduces reliance on fully labeled datasets by leveraging partially sampled k-space data, thus making advanced MRI reconstruction more accessible across varied clinical and research settings.
Further, the project investigates federated learning approaches to enable multi-site collaboration while preserving patient privacy. This ensures more generalized and reliable reconstruction models that can adapt to diverse data distributions. Additionally, we explore dynamic MRI reconstruction by learning spatiotemporal correlations, enhancing applications such as real-time imaging and cardiac MRI. Whether you are intrigued by applying generative models like GANs and diffusion probabilistic models or developing joint reconstruction frameworks for multi-modality imaging, this project offers a diverse and impactful research scope. Aspiring students will have the opportunity to contribute to real-world healthcare advancements and innovative AI solutions.
Required knowledge
1. Foundational Understanding of AI and Machine Learning:
- Knowledge of machine learning algorithms, optimization techniques, and deep learning frameworks (e.g., PyTorch, TensorFlow).
- Familiarity with neural network architectures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs).
2. Signal and Image Processing Basics:
- Understanding of Fourier transforms, sampling theory, and image reconstruction concepts.
- Experience with medical imaging or related fields is a plus.
3. Programming Skills:
- Proficiency in Python for AI/ML development.
- Familiarity with scientific computing libraries (e.g., NumPy, SciPy) and visualization tools (e.g., Matplotlib, Seaborn).
4. Knowledge of MRI Physics and Acquisition Principles (optional but advantageous):
- Basic understanding of k-space, MRI signal acquisition, and reconstruction pipelines.
5. Problem-Solving and Research Aptitude:
- Ability to work with complex datasets and debug computational issues.
- Critical thinking to contribute to innovative research directions and adapt existing models to MRI data.
6. Collaboration and Communication Skills:
- Willingness to collaborate in interdisciplinary settings, bridging AI and healthcare domains.
- Ability to document research progress and communicate findings effectively.
