Artificial intelligence and specifically federated learning can improve the detection of brain tumours by as much as 33%. This is according to a joint study Intel Labs and the Perelman School of Medicine at the University of Pennsylvania (Penn Medicine). The Dutch Erasmus MC in Rotterdam is also involved in the project.
Figures from KWF Kanker Bestrijding show that a total of 1,397 Dutch people suffered a brain tumour in 2021. 824 people died of a brain tumour in 2021. The survival rate over a five-year period is 28% and over a 10-year period 21%.
Federated learning
The study focused on federated learning, a form of artificial intelligence (AI) based on distributed machine learning (ML). "Federated learning has huge potential in numerous domains, especially in healthcare, as evidenced 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 around the world, and we look forward to further exploring the promise of federated learning," explains Intel Labs chief engineer Jason Martin.
The parties investigated the use of federated learning for the identification of malignant brain tumours. This used very large dataset from 71 institutions on six continents. Using such a dataset is complex, partly because of strict national data protection laws. Consider the General Data Protection Regulation (AVG) in the Netherlands.
Stringent legislation makes it difficult to achieve medical research and data sharing on a large scale. Federated learning could offer a solution. The parties report that Intel's federated learning hardware and software meets data privacy requirements and protects data integrity, privacy and security through confidential computing.
Decentralised system
Data was processed in a decentralised system using Intel federated learning technology. This combined with Intel Software Guard Extensions (SGX). The technology ensures that the raw data stays within the hospital's own network. Based on the raw data, the technology calculates model updates. Only these model updates are forwarded by the technology to a central server or aggregator. In this way, the technology addresses various data privacy issues.
Radiologist Prof Dr Smits and biomedical researcher Dr Van der Voort from Erasmus MC explains: "From Erasmus MC, this federated learning study allowed us to contribute to improving automatic tumour detection without having to send patient data. Automatic tumour detection is an important step for personalising and monitoring treatment, and to develop this methodology it is essential to use data from many different institutes. With this collaboration, we were able to do that easily, while keeping control of our data ourselves."
Secure multi-institutional collaborations
"Federated learning offers a breakthrough in ensuring secure multi-institutional collaborations. It enables access to the largest and most diverse dataset ever seen in the literature. And all this while keeping all data within each institution itself at all times," said senior author Spyridon Bakas, PhD, assistant professor of Pathology & Laboratory Medicine, and Radiology, at the University of Pennsylvania's Perelman School of Medicine. "The more data we can feed into machine learning models, the more accurate they will become. That, in turn, will improve our ability to understand and treat even rare diseases, such as glioblastoma."
The results of the study have been published in the scientific journal *Nature Communications*.
Author: Wouter Hoeffnagel
Photo by: Pixabay / Michal Jarmoluk
