publive-image

The healthcare industry is the most crucial industry that relies entirely on vast sets of data to deliver quality healthcare. This healthcare data is quite valuable and vulnerable as well to data theft. Machine learning models promise the security of patient data, but the traditional ML techniques can pose a risk to patient data privacy. This is where federated learning comes in, bringing ML models to the data source, instead of bringing the data to the model. It ensures the security of user’s data by enabling their data to stay on the device where they reside. It also enables the applications running a specific program are still learning how to process the data and creating a better, more efficient, model.

Federated learning trains machine learning algorithms on private, fragmented data, stored on various servers and devices. Initially developed for diverse domains, such as mobile and edge device use cases, it has recently gained more interest in healthcare. But the question raised why federated learning is important for healthcare applications despite having machine learning models?    

One of the reasons is IBM’s Watson, the popular AI application in healthcare. In 2017, the systems reportedly prescribed a drug that could have claimed a patient’s life during a simulation. This was happened owing to poor training of Watson’s software, according to internal IBM documents, saying it only used a limited set of hypothetical cancer cases, rather than real patient data. Additionally, it followed recommendations from just a few specialists instead of verified guidelines or evidence. However, it demonstrated that despite the promise of AI to transform healthcare, the ability to access and use patient data in the right ways is hugely challenging for a traditional machine learning model.

Federated Learning to Transform Healthcare

Research has revealed that federated learning-enhanced models can achieve performance levels against those trained on centrally hosted data sets and superior to models that only see isolated single-institutional data. The implementation of federated learning could also enable precision medicine, leading to models that yield unbiased decisions and are sensitive to rare diseases while concerning governing and privacy concerns.

Companies like IBM Research, Owkin and others, are capitalizing on federation learning to address challenges in healthcare. Owkin, a French-American AI-powered life science company, uses federated learning to accelerate medical research. The company’s use of this model helps deliver effective treatment and drugs to cancer patients. The startup is collaborating with the U.S. and European cancer centers to use their data for its models. It is also developing a new model that foresees survival odds for a rare type of cancer-based on a patient’s pathology images.

On the other side, GPU maker NVIDIA at RSNA introduced a solution, dubbed NVIDIA Clara Federated Learning. It uses a distributed, collaborated learning technique that keeps patient data inside the walls of a healthcare provider. Clara Federated Learning runs on NVIDIA EGX intelligent edge computing platform. It is a reference application for distributed, collaborative AI model training that preserves patient privacy. Even though federated learning is relatively new, it has proved its validity in securing patient data. It still requires technical consideration to ensure that the algorithm is proceeding optimally without compromising safety or patient privacy.