It takes two: ‘No clinical AI solution can work on its own’

By Melissa Fassbender

- Last updated on GMT

(Image: Getty/ipopba)
(Image: Getty/ipopba)

Related tags AI Clinical trial Real world data Real world evidence

Clinical AI, no matter how sophisticated, cannot work on its own, says Life Image CEO, who predicts that CROs will either adapt or be disrupted by smaller companies with digital access capabilities.

Life Image has the world’s largest interoperable global network for accessing data and clinical endpoints, connecting 1,500 US hospitals with 150,000 providers and 58,000 global clinics to manage 10m clinical encounters per month, According to the company.

Its network collates an “ecosystem”​ of hospitals, clinicians, patients, care teams, life sciences, medical devices, telehealth, and electronic health records (EHRs). And with the ability to scale up to a population health level, CEO Matthew Michela said it can help advance clinical trials, incorporating imaging data into real world evidence (RWE) programs in life sciences.

As part of this, Life Image recently announced a strategic partnership with Mendel Health, a San Francisco, CA-headquartered company which uses artificial intelligence (AI) to examine unstructured data in medical literature and EHRs.

Through its partnership with Mendel, Life Image is using the platform to facilitate clinical trial site location as well as patient recruitment for oncology studies.

According to the company, Mendel Brain yields 30 to 50% more patients than traditional methods, using machine learning (ML) to process information and continuously improve outcomes to decrease the time between patient eligibility and identification.

“What makes the solution work in practice is connecting the technology to medical data, which requires both platform and clinical workflow integration that is supported by Life Image across its large network of hospitals and clinical trial sites,”​ Michela explained.

Mendel Brain also is focused specifically on oncology, which Michela said allows it increase precision in a therapeutic area. “The more general an AI, the poorer its performance will be when faced with any specific complex problem,”​ he explained.

Michela noted that creating innovative AI is a challenge – and up to 70% of AI products fail​, according to a panelist during the SCOPE Summit in February.

“Many of these types of solutions do not get put into practice because in order to actually work, they require integration into a wide array of data storage sites across a wide set of hospitals and clinical trial sites, all of which have different technology platforms,” ​Michela added.

Life Image addresses this challenge as it has built interoperable integrations into thousands of hospitals that support millions of patients monthly, across all types of physician workflows, he explained.

Still the need for scalable technology solutions integrated into the provider and hospital workflow will accelerate, as the amount of data sources continues to grow rapidly.

Among these sources is image data, which Michela predicts will “rapidly start”​ to be included in real world evidence (RWE), which he said will develop into the primary way clinical trials are conducted in both the pre- and post-markets.

He also predicts that contract research organizations (CROs) and the heavy manual process of collecting data will either adapt or be disrupted by smaller companies with digital access capabilities.”

Michela said, “No matter how sophisticated it is, no clinical AI solution can work on its own. It needs access to massive amounts of interoperable data at scale within the delivery system in order to have any utility.”

Related news

Show more

Related products

show more

The Right CDMO Can Unlock Faster Drug Development

The Right CDMO Can Unlock Faster Drug Development

Content provided by Lonza Small Molecules | 15-May-2023 | Interview

Drug development presents biotech companies with ever-changing challenges. Enabling a faster end-to-end process requires seamless flexibility and extensive...

SDTM supplemental qualifiers explained

SDTM supplemental qualifiers explained

Content provided by Formedix | 12-May-2023 | White Paper

What are SDTM supplemental qualifiers? In short, these are variables in non-CDISC datasets that cannot be mapped to a variable that matches the SDTM standard....

Plan for success with process liquid and buffer preparation

Plan for success with process liquid and buffer preparation

Content provided by Thermo Fisher Scientific - Process Liquid Preparation Services | 25-Apr-2023 | White Paper

Setting the groundwork for successful scale-up is essential for getting a therapeutic to market quickly and efficiently, but navigating the unknowns associated...

Related suppliers

Follow us


View more