Leukemia generally affects the bone marrow and blood, which is a need for early detection and precise treatment planning. The advancement of diagnostic models in medical imaging impacts patient outcomes. However, traditional approaches to model training create substantial privacy risks due to the sensitive patient data. FL appears to be a viable concept that protects patient privacy while facilitating cooperative training of models across decentralized sources of data. This method complies with ethical standards and privacy regulations, mitigating risks of data breaches and unauthorized access. Federated learning advances the collective intelligence of diverse medical imaging datasets from multiple institutions. By integrating federated learning principles, researchers can achieve advancements in leukemia detection while keeping high standards of data confidentiality. The Fed AVG algorithm shows us this approach, demonstrating efficient, privacy-preserving model training through iterative local updates and global aggregation.