Comunicado de prensa
Intel Labs and the Perelman School of Medicine at the University of Pennsylvania (Penn Medicine) are working together to develop a technology that will allow a network of 29 health centers and research institutes led by Penn Medicine to collaborate with Artificial Intelligence (AI) types brain tumors using a secretive process called Federal learning.
Penn Medicine's work will be funded by the National Cancer Institute (NCI )'s National Cancer Institute (NIH) Cancer Research Information Technology (NIH) program, with a three-year grant of $ 1.2 million awarded by a senior investigator, Dr. Spyridon Bakas, of the Center for Computing and Analytics of Biomedical Images (CBICA) of the Faculty of Medicine Perelman of the University of Pennsylvania.
"Artificial Intelligence is very promising Early detection of brain symptoms, but to reach their full potential, requires more than one medical center. Through Intel's software and hardware, which is supported by some of our brightest company ideas, we are partnering with the University of Pennsylvania and the coalition of 29 medical centers to promote identification of brain signals while protecting the brain. sensitive patient information. ”-
Jason Martin, Senior Engineer at Intel Labs How It Works: Cooperative Learning, which is a distributed machine learning system, will be used by Penn Medicine and 29 health and research centers in the United States, Canada, United Kingdom, Netherlands, Germany, Switzerland and India. allowing organizations to collaborate on deep learning projects without sharing patient data.
Penn Medicine and Intel were the first to publish a document on integrated learning in the field of medical imaging, showing in particular that the method used could train the model with 99% more accuracy compared to the trained model in the traditional way. , not in private. This document was first presented at the International Conference on Medical and Computer Aided Intervention (MICCAI) 2018, Granada, Spain. The new project will use Intel and Hardware software to apply federal education in a way that offers greater protection of privacy in the model and data.
"Our scientific community accepts that machine learning requires large quantities and variations of data that no institution has. We are coordinating a network of 29 international health and research centers that will be able to train health AI models using state-of-the-art machine learning techniques, including integrated learning. will begin developing advanced brain tumor identification techniques from an expanded version of the data set for the International Brain Tumor Segmentation Challenge (BraTS) .These features will allow medical researchers access to large amounts of health data, while protecting the safety of that information. ”- Dr. Spyridon Bakas of the University of Pennsylvania
Why it matters
According to the American Brain Tumor Association (ABTA), about 80,000 people will be diagnosed with a brain tumor this year, and more than 4,600 of them will be children. In order to train and build a model for finding a brain tumor, which can help early detection and better outcomes, researchers need to access large amounts of relevant medical data. However, it is important that data is not confidential and protected, and that is where technology comes into the organization's learning. Using this approach, researchers from all partner organizations will be able to work together to develop and train an algorithm to detect brain tumors while protecting sensitive medical data.
By 2020, Penn and 29 global health and research centers will use Intel's study equipment to produce the new state-of-the-art Artificial Intelligence, which is being trained with the largest data sets of the brain so far. date – all without confidential patient confidentiality. The subsidiary of the affiliated institutions expected to participate in the first phase of this collaborative effort includes, but is not limited to, Pennsylvania University Hospital, Washington University of St. Louis, University of Pittsburgh Medical Center, Vanderbilt University, Queen & # 39; s University, Technical University of Munich, University of Bern, King & # 39; s College London and Tra Memorial Hospital.