Intel and Penn Medicine Announce Results of Largest Medical Federated Learning Study :
Using federated learning, a distributed machine learning (ML) artificial intelligence (AI) approach, Intel Labs and the Perelman School of Medicine at the University of Pennsylvania (Penn Medicine) have finished a joint research study to assist international healthcare and research institutions identify malignant brain tumours. The experiment showed the ability to increase brain tumour identification by 33% and was the largest medical federated learning study to date with an unparalleled global dataset reviewed from 71 institutions across six continents.
“Federated learning has tremendous potential across numerous domains, particularly within healthcare, as shown by our research with Penn Medicine. Its ability to protect sensitive information and data opens the door for future studies and collaboration, especially in cases where datasets would otherwise be inaccessible. Our work with Penn Medicine has the potential to positively impact patients across the globe and we look forward to continuing to explore the promise of federated learning.”
Jason Martin, principal engineer, Intel Labs
What Is Important:
Because of state and federal data privacy rules, notably the Health Insurance Portability and Accountability Act, data accessibility has long been a problem in the healthcare industry (HIPAA). As a result, it has been extremely difficult to conduct large-scale medical research and data sharing without jeopardising patient health information. In order to protect data integrity, privacy, and security through confidential computing, Intel’s federated learning hardware and software complies with data privacy concerns.
With the help of Intel federated learning technology and Intel® Software Guard Extensions (SGX), which eliminate data-sharing barriers that have traditionally prevented collaboration on similar cancer and disease research, the Penn Medicine-Intel result was achieved by processing large volumes of data in a decentralised system. By keeping raw data within the computational infrastructure of the data holders and only enabling model updates generated from that data to be transferred to a central server or aggregator, the method addresses many data privacy problems.
“All of the computing power in the world can’t do much without enough data to analyze”, “This inability to analyze data that has already been captured has significantly delayed the massive medical breakthroughs AI has promised. This federated learning study showcases a viable path for AI to advance and achieve its potential as the most powerful tool to fight our most difficult ailments.”
Rob Enderle, principal analyst, Enderle Group
“In this study, federated learning shows its potential as a paradigm shift in securing multi-institutional collaborations by enabling access to the largest and most diverse dataset of glioblastoma patients ever considered in the literature, while all data are retained within each institution at all times. The more data we can feed into machine learning models, the more accurate they become, which in turn can improve our ability to understand and treat even rare diseases, such as glioblastoma.”
Spyridon Bakas, PhD, assistant professor of Pathology & Laboratory Medicine and Radiology at the Perelman School of Medicine
Researchers need to have access to vast quantities of medical data to enhance the treatment of diseases; in most cases, these datasets are larger than what a single facility can produce. The study shows the efficacy of federated learning at scale and the potential advantages multisite data silos can uncover for the healthcare sector. Early disease detection has advantages because it may prolong a patient’s life or improve their quality of life. The findings of the study conducted by Intel Labs and Penn Medicine were released in the peer-reviewed journal Nature Communications.
Concerning the Research:
Glioblastoma (GBM), the most prevalent and fatal adult brain tumour with a median survival of just 14 months after standard treatment, is a rare form of cancer. In 2020, Intel and Penn Medicine announced an agreement to collaborate and use federated learning to improve tumour detection and treatment outcomes. Over the past 20 years, treatment options have increased, but overall survival rates have not increased. The National Cancer Institute of the National Institutes of Health’s Informatics Technology for Cancer Research programme provided funding for the study.
To enhance the detection of rare cancer borders, Penn Medicine and 71 international healthcare/research institutes deployed Intel’s federated learning hardware and software. Radiologists used a brand-new cutting-edge AI software platform called Federated Tumor Segmentation (FeTS) to define a tumor’s perimeter and enhance the recognition of the “operable region” or “tumour core” of tumours. Radiologists annotated their data and ran the federated training using open federated learning (OpenFL), an open source framework for training machine learning algorithms. The platform was trained using the largest brain tumour dataset to date, consisting of 3.7 million pictures from 6,314 GBM patients spread across six continents.
By working together on this project, Intel Labs and Penn Medicine have produced a proof of concept for federated learning as a method of learning from data. The answer may have a big impact on healthcare and other fields of study, including different kinds of cancer research. For example, Intel created the OpenFL open source project to help clients embrace cross-silo federated learning in the real world and reliably implement it on Intel SGX. In order to facilitate ongoing development and promote collaboration between the FeTS platform and Intel’s OpenFL open source toolkit, both of which are accessible on GitHub, the novel FeTS initiative was also established as a collaborative network.