Revealing Brain Tumor with Federated Learning

Brain and other nervous system cancer ranks as the tenth most common cause of death. According to estimates, primary cancerous brain and central nervous system tumors will be the cause of 18.990 deaths in the United States in 2023, more precisely 11.020 men and 7.970 women. Primary cancerous brain and central nervous system tumors are estimated to have killed 251.329 people globally in 2020. Given the variety and potential danger, it becomes evident that it is necessary to promptly detect any abnormal formations within the head in order to proceed with timely therapy. In this regard, artificial intelligence has proven to be useful for constructing systems capable of detecting, from medical images, the presence of cancerous formations in a non-invasive and timely manner. Questions regarding how data are handled always spark discussions on ethical grounds and on privacy protection. Federated Learning is a specific protocol that manages data locally within the involved medical institutions, aiming to address privacy-related issues. The objective of the paper is to assess the existing state-of-the-art solutions by emphasizing their workflows and key components. This analysis takes into account the datasets employed and the architectural decisions made. Finally we scrutinize open challenges and potential future development to address these specific issues.