MIT, Philips Expand Dataset for Health AI Development

Massachusetts Institute of Technology (MIT) and health technology company Philips announced an expansion to their critical care dataset at HIMSS 2023 earlier this week, giving researchers access to clinical data from 200,000 patients to help advance artificial intelligence (AI) and machine learning (ML) in healthcare.

The dataset, eICU Collaborative Research Database (eICU-CRD), is an expansion of an initiative between Philips and MIT’s Institute for Medical Engineering and Science (IMES) that aims to improve patient care and clinical outcomes through the development of AI-based solutions.

The dataset was first released in 2016, and includes de-identified information such as diagnoses, lab results, severity of illness scores, vital signs, and pharmacy and medication orders for patients from 200 hospitals in the US. The press release states that these data can then be used to generate insights into treatments, readmissions, comorbidities, and clinical outcomes.

Care quality and delivery challenges stemming from the COVID-19 pandemic led Philips and IMES to expand the dataset.

“The database, which includes patient information from 2020 and 2021, now contains significant overlap with the Covid-19 pandemic, yielding valuable patient data for research,” said Leo Anthony Celi, MD, principal research scientist and clinical research director at the Laboratory of Computational Physiology at IMES, in the press release.

The database also has the potential to bolster medical and informatics training.

“This updated database is a vital resource for education, including in many courses at institutions like Harvard, MIT and Stanford; and training, as well as low-resource institutions,” said Jesse D. Raffa, PhD, research scientist in the Lab for Computational Physiology at IMES.

Credentialed medical researchers who complete human subjects training and agree to the data use agreement will be granted access to the data by Philips and IMES, according to the press release.

The Laboratory of Computational Physiology will serve as the academic research hub for the initiative, providing and maintaining access to the database. The hub will also help educate researchers on the database and offer a collaboration platform.

“This [initiative] is how we can enhance patient care and improve clinical outcomes: liberating and connecting data across systems and applications with integrated devices, systems and informatics, which can inform research with patient insights that can help clinicians make the right decision at the right time for their patients,” said Shiv Gopalkrishnan, general manager of EMR & Care Management at Philips.

MIT and Philips have both made additional efforts to develop AI-based tools for healthcare in recent years.

In January, researchers from MIT and Mass General Brigham shared that they developed a deep learning tool capable of predicting individual lung cancer risk within six years for patients with or without a significant smoking history using low-dose chest computed tomography.

In 2020, Philips partnered with the US Defense Threat Reduction Agency (DTRA) and the Defense Innovation Unit (DIU), parts of the US Department of Defense, to expand an AI-based early warning system to monitor and contain COVID-19.