Where AI Can Add Efficiencies To Clinical Trials
By Deborah Borfitz
January 20, 2022 | Acceptance of artificial intelligence (AI) in clinical research is on the rise thanks, in part, to its pairing with human expert oversight, according to Kevin Thomas, Ph.D., co-director of AI at Clario, the technology company born of the recent merger of ERT and Bioclinica. The result is faster, more efficient clinical trials with less of the bias associated with purely manual evaluations while sidestepping the potential risks of AI.
AI-enabled applications at Clario are designed to “help prevent humans from errors or creating errors,” adds Janine Jones, senior product manager for adjudication and eligibility. “Anywhere we can implement AI to do that is what we want to do.”
Those efforts picked up steam earlier last year when then-Bioclinica acquired Saliency and began integrating the company's advanced AI technology into its imaging platform, says Jones. Thomas was at the time serving as CEO of Saliency.
Thomas and Jones will be presenting on the use of AI to improve clinical trial efficiencies on Feb. 10, the final day of the 2022 Summit for Clinical Ops Executives (SCOPE). Together with Łukasz Kidziński, Ph.D., co-director of AI at Clario, their talk will cover imaging and source document redactions as well as electronic clinical outcome assessment (eCOA) in decentralized clinical trials.
A lot of attention has been paid to protecting personal health information (PHI) because a breach could be potentially catastrophic, she notes. If a human misses PHI in the redaction process, the study sponsor, the CRO, and the software provider could be penalized millions of dollars—not to mention the incalculably high cost of a bad reputation.
AI is already being used to detect possible PHI in text information captured in clinical trials, including identifiers such as date of birth and medical record number referenced in physician notes and PDFs (e.g., lab values, discharge summaries, and CT scan results). The solution scans documents and flags all identified PHI, but people doing the redaction independently decide what does and does not get redacted, Jones says.
One major use of AI currently is in support of the image redaction process in studies involving video footage of study participants, both to protect patient privacy and eliminate the time-intensive step of manually obscuring their faces, Thomas adds. “Even if we have a human that is double-checking the whole video, it is a lot faster to check the AI’s work than to be going in and blurring out the faces in every video frame.”
Along the same lines, Thomas says, Clario will soon be rolling out software in a clinical trial to blur faces in brain MRIs while maintaining the integrity of the imaging data for evaluating the health of study participants. As reported a few years ago by the Mayo Clinic, commercial facial recognition software can be used to identify people from brain MRIs that includes imagery of the face, despite steps that researchers typically take to protect patient privacy.
In The Queue
Clario is also actively working on solutions using AI to automate the laborious image preparation process, says Thomas. A great deal of work is involved in taking a picture captured by an X-ray or MRI machine and getting it ready for a radiologist to read, including being sure it is a high-quality image and properly formatted and rotated.
“Even in the very rare instance where AI might make a mistake in that process, it wouldn’t jeopardize a clinical trial,” Thomas says. “It is almost purely adding benefit without any risk.”
Over the next three to five years, Clario expects to introduce tools using AI-driven predictive analytics, he continues. This would include biomarkers of potential future risk or areas on CT images of a tumor that might warrant greater clinician attention. “CT scans are very large files comprised of many slices depicting a full region of the body, so there is always a potential risk of overlooking something.”
Solutions that tailor patient engagement to improve protocol compliance is a nearer term ambition, Thomas says, notably to reduce the burden of participation in decentralized clinical trials. While decentralizing study features makes it easier for patients by limiting how often they must travel to a study site, doing tests and completing surveys at home on their own requires more engagement and mental effort from them than a traditional clinical trial.
“There are many reasons to be excited about the potential of decentralized clinical trials and remote health monitoring, but we need to be careful about how much we are asking of patients,” says Thomas. “We’re constantly looking for ways to facilitate more intuitive home health assessments to balance the need for data with the need to respect patients’ time and privacy.”
It might be that a prognostic variable is continuously measured to relieve participants of the inconvenience of having to periodically repeat a test or the user interface is designed to minimize time spent entering information, he offers as examples. At-home monitoring devices might also be more intuitive to operate, to save patients the hassle of having to travel to a medical facility to have tests done.
Clario views AI as a tool to be incorporated into its Trial Anywhere solutions portfolio wherever it can facilitate participation in clinical trials while also ensuring the data produced are of high quality and align with the study protocol, Thomas says. The company’s advantage comes from the combined technology and scientific expertise of ERT and Bioclinica, which united has been in business for 50 years and supported 19,000 clinical trials—including 70% of studies conducted between 2019 and 2020.
Interest in decentralized trials is at an all-time high, says Thomas. One of the key issues here has been getting disparate and disconnected datasets talking to one another.
“Now that the two legacy companies have come together, much of our focus is on unifying many different clinical endpoints,” he says. A large, complex trial might create data streams from at-home wearable device recordings, images collected at satellite imaging centers, and clinical evaluations done at bricks-and-mortar sites, and Clario aims to merge that information into a single portfolio.
Decentralized clinical trials encompass multiple categories of clinical endpoints and strategies for assessing patients’ health closer to home, says Thomas. One example is the introduction of satellite imaging centers to save participants the time it takes to travel to a traditional trial site further from their home.
Equally important are wearables, including smartwatches and accelerometers that measure patients’ movement patterns and attach to different parts of their body, he continues. The information is often collected on patients with either neurological or musculoskeletal conditions but can also be useful for quality-of-life assessments for people with cancer.
A third category is eCOA, including patient-, clinician-, and caregiver-reported outcomes that can be collected at home with optional online interaction with a physician, Thomas says. “Clario is very excited to bring all of these endpoint solutions together to streamline decentralized trials.”