By Kevin Davies
April 4, 2012 | It has taken more than a decade, but the ideas of Venkatesan Thangaraj to create a pharma pipeline for image management are maturing so quickly that he sees the potential of developing an industry standard that enables all trial sponsors and vendors to have a single customizable tool for handling clinical imaging data.
Thangaraj founded Radiant Sage Ventures in 2008, hoping to offer pharma companies “a missing link” in their imaging operations. “Around then, there was much ado about imaging,” he says. “Pharma was putting out RFPs to build or buy solutions, but there just weren’t any. Many pieces went back in house.”
But Thangaraj’s ideas around imaging software dated back to 1999, when he first put together a business plan for such a system. Back then, however, VC firms he pitched wanted to focus on EDC rather than imaging. “I spent a good part of 7-8 years working on EDC technologies,” he says. (He founded ENMED in 1999, and eventually sold it to an Indian IT company called MegaSoft.)
So when Thangaraj heard of pharma’s renewed interest in imaging software, he already had the core concepts around what he wanted to do. It was just a matter of implementing it. In 2008, he arranged to meet with Pfizer, with which he had worked on other projects, to discuss his ideas and dust off his old slides. Six months after he built a prototype system, Pfizer became his first client, commissioning what became the Image Collaborative Portal (ICP).
Kris Kokomoor, associate director of Pfizer’s Center for Excellence for Drug Development group in Groton, Connecticut, says Pfizer enjoys a strong relationship with Radiant Sage and continues to pursue the same goals in managing clinical image data. “We’re at the same spot on the curve as several other big pharmas,” he says. “Other pharmas might try to solve this problem internally, but I’m glad we didn’t do that.”
“One of our goals,” he continues, “was to procure a product that would gain support in other areas of the industry. We’ve encouraged Radiant Sage to work with other pharmas. Most [software] solutions are very proprietary, and that’s not a direction we’re heading in.”
From Pfizer’s perspective, Kokomoor says there was a gradual recognition dating back five years or more that it was leaving good data in the field. “There was no epiphany—or if there was, it certainly preceded me!”
“We’d had many studies incorporating clinical imaging end-points, but the images were typically used only to derive endpoint data,” he says. “The images were often left at the acquisition site or CRO. They’d read the images in search of a clinical endpoint.”
While the clinical team might be satisfied with the endpoints, other analyses would require Pfizer researchers go back to the CRO on the original trial and ask for a copy of the images. Those images would typically be dispatched on DVDs or CDs from hospitals or CROs, their receipt tracked on Excel spreadsheets. The researchers typically lost interest, Thangaraj says, because they couldn’t receive information in a timely fashion. “Every trial was different,” recalls Kokomoor. “We had to deal with them ad hoc… What we wanted was a toolbox to transform image sets for new use.”
The first version of ICP went live in December 2009, and provided an archive where live clinical trial data could be imported and indexed. “Anyone else could search the archive and export it out in the next day or two,” Thangaraj recalls.
By imposing a process and consistency on the front end while importing image data into Pfizer, Kokomoor and his colleagues reckoned they could reduce the amount of time servicing requests for further use of the data. Most imaging results are obtained in a DICOM file format, which defines a lot of metadata. “Those data contain a rich set of information, but some aspects are common and standardized, others aren’t. We try to impose some standardization across study boundaries, so cross-study queries are effective.”
Radiant Sage designed a web-based app with several Java-based applets for data import. “We provide the access to the app and the training [the CROs] need to get the job done,” says Kokomoor. That extends to reading the images. “We provide training for readers, consumers, or importers.”
The major return-on-investment is in the time saved between sending a request for clinical data and when the data are available. “We’ve made real progress in minimizing that amount of time. It’s a big-time savings,” says Kokomoor. “I’m confident this is adding to our ability to remain compliant, [improve the] chain of custody for images, ensuring images in house are properly de-identified. Now we retain the documentation.”
In one example, many Pfizer clinicians wanted to study MRIs and X-rays associated with an osteoarthritis study involving 400 subjects over four years. “Any one data extract could have required months of manpower—now it’s reduced to man hours or days to get done,” says Kokomoor.
Kokomoor stresses the continued malleability of Radiant Sage and its willingness to build to Pfizer’s requirements. Part of the original specification was the creation of a repository within Pfizer’s datacenter to hold terabytes of indexed image data and render it accessible in various ways. “It’s like a PACS—but has a more study-centric view,” he says.
The network-attached storage system manages tens of terabytes of data. Privacy concerns led Pfizer to choose an internal solution rather than house the data externally, at least for the time being. “The landscape around cloud computing—putting sensitive data in secure repositories outside the corporate firewall—that’s changed a lot in the past two years. That’s an aspect that we’ll be revisiting over time,” says Kokomoor.
Early on, the IT infrastructure included Nvidia GPUs, but more recently, Radiant Sage has revised the infrastructure to use just CPU power. Pfizer IT heavily leverages a shared infrastructure, Kokomoor explains, and that wasn’t helped by using a specialized advanced processing unit such as Nvidia.
Kokomoor says a major disease area that has benefited from the relationship is neuroscience—acquiring images of brains. Quite frequently, clinicians want to review the primary image data to check for artifacts or verify a set of screening decisions. Now the Imaging CRO can transfer the images to Pfizer, where staff clinicians can review the images.
Radiant Sage has also introduced remote rendering for data. This allows a clinician to pull up a web tool and browse the data without download of any images. “We wanted them to view—so they implemented remote rendering, back on the client’s system,” says Kokomoor.
Last year, Radiant Sage was awarded projects by another top 5 pharma, which involved core lab work. “We call version 2.0 a Core-Lab-in-a-Box,” says Thangaraj. It added workflow features, radiology reads, and integration with Electronic Case Report Forms (eCRFs).
Thangaraj says he discussed imaging software with another large pharma five years ago. “They built an in-house solution; it’s their internal archive. But other features are not part of it, so I went back to them, showed them what we had, and there was some interest to look beyond what they’d built.” Radiant Sage recently won a project to take the open-source ImagEDC code and validate it to make it 21 CFR Part 11 compliant.
Radiant Sage also recently expanded its business model to provide the same kind of solution implemented by Pfizer as a service. “The use case hasn’t changed, but we’re getting into more collaborative research with external parties,” says Thangaraj. “Can imaging help us? Who is the knowledge expert? Who is using imaging? The trend is to collaborate or provide funding to these researchers.”
Thangaraj offers an example: “If we have a site in Philadelphia that does some analysis, they can send this to another researcher to re-analyze the experiment. Now we have two independent sources of results. This ad hoc research happens frequently. Medical imaging plays a strong role. Over the last few months, that’s the focus of our discussions. How do we get collaborators onto a common platform so straightforward to start to look and share images?”
“We want to stay focused on technology, but our partnership with Impact CoreLab enables us to provide services around our technology,” says Thangaraj. “We can provide customers a one-stop shop from early research to clinical trials.”