Medical image analysis is rapidly advancing, driven by machine learning and its potential to improve diagnosis, treatment planning, and disease understanding. However, training robust and generalizable AI models in this domain requires access to large, diverse datasets of medical images. Due to stringent patient privacy regulations and data sensitivity, directly sharing medical image data across institutions is often restricted or impractical. This project addresses the critical need for privacy-preserving medical image analysis by developing a federated learning framework that enables collaborative model training without compromising patient confidentiality. This project will focus on designing, implementing, and evaluating a federated learning system and compare it with its centralized counterpart for medical image analysis. The core problem is to enable multiple medical institutions to collaboratively train a shared machine learning model on their respective image datasets, without directly exchanging or centralizing the sensitive patient data. This project will investigate the technical challenges of federated learning in the medical imaging context, such as handling non-identically distributed (non-IID) data across institutions, and managing communication efficiency. The goal is to create a functional and robust federated learning framework that demonstrates the feasibility and benefits of privacy-preserving collaborative AI in medical imaging. The project will involve several key development and research tasks: (1) Implementing a federated learning framework using a suitable platform (e.g., Flower, TensorFlow Federated) and adapting it for medical image data. (2) Quantify the performance trade-offs between centralized and federated learning approaches . (3) Evaluate the impact of data heterogeneity on federated model convergence and performance. (4) Assess architecture robustness in a federated setting. (5) Quantify the implicit regularization effects of federated averaging. Expected outcomes include a functional federated learning framework for medical image analysis, a comprehensive evaluation of its performance, a comparison of federated model with its counterpart, and a clear analysis of the challenges and opportunities of applying federated learning in this sensitive domain. The aim is to provide a valuable open-source resource and contribute to the advancement of AI in healthcare. Students undertaking this project should possess strong programming skills and a solid understanding of machine learning principles, particularly deep learning. Proficiency in Python and experience with deep learning frameworks such as TensorFlow or PyTorch are essential. Familiarity with distributed systems, networking concepts, and ideally, some background in medical image analysis techniques would be highly beneficial. Access to computational resources, including GPUs for model training, will be necessary and university lab resources can be utilized. The code developed within this project is expected to be released as open-source software under the AGPL v3 or a compatible license.