How Did Clinical Researchers Actually Use AI In 2024?
As told to Abby Proch, executive editor of guest columns, Clinical Leader

Ask around, whether at home or in the workplace, and you’ll find that nearly everyone has heard about AI and all its touted benefits. But dig deeper with an albeit generic follow-up question — “Have you ever used AI before?” — and you might get more hems and haws. Generally speaking, about one in three people have used a generative AI chatbot. Even fewer have likely tapped into AI’s many uses beyond a glorified Google search.
But as 2024 comes to a close and we welcome 2025 and all the advancements it may bring with the inclusion of AI in everyday life and work, Clinical Leader sought to understand more specifically how clinical research professionals are incorporating AI into their work.
We asked, “Simply put: Are you using any iteration of AI in your role within clinical research? If so, how and why?”
Below, you’ll find a handful of ways in which industry and academia are embracing the use of AI, from drug discovery and development to niche research projects.
"At Houston Methodist, we integrate AI with deep biological and clinical expertise to revolutionize clinical research, streamline processes, reduce costs, and improve patient outcomes. Our peer-reviewed AI applications span multiple clinical disciplines. We take a comprehensive, enterprise-wide approach by embedding AI seamlessly into clinical workflows. “For example, ischemic stroke misdiagnosis in emergency departments in the United States exceeds 22%, particularly in mild to moderate cases where symptoms often overlap with other conditions. To address this, we mapped the entire stroke workflow in the emergency department and pinpointed critical pain points. “In response, we developed DeepStroke, an AI tool that analyzes facial videos and speech to enhance stroke triage accuracy. Additionally, we designed a mobile imaging device incorporating AI-driven CT denoising and net water uptake algorithms — key components for improving stroke and edema detection. Clinical validation showed significant improvements in stroke triage and diagnosis accuracy. Both technologies have successfully undergone clinical evaluation and are currently being assessed in prospective clinical trials at the hospital, focusing on stroke-susceptible patients arriving at the emergency room, including underserved populations that are primarily Spanish-speaking. We are happy to share our AI technology and experience with other hospital systems to improve stroke care.” |
|
“We recently used artificial intelligence in our trial data analysis to enhance our data interpretation capabilities, addressing a key issue: high variance in trial outcomes, likely due to patient heterogeneity. Specifically, we applied a machine learning-based model to address high variance in our trial outcomes. This high variance, attributed to diverse patient responses, diminished the study’s statistical power and hindered our ability to clearly demonstrate treatment efficacy. The ML model leveraged collected patient data such as demographic and baseline health metrics to explain the variance introduced, effectively isolating the true treatment effect from background noise. “By accounting for these patient-specific factors, the Placebell method (offered by Cognivia) reduced unexplained variance, enabling us to detect efficacy signals masked in previous analyses and reach statistical significance without repeating the trial.” |
Arjan Blokland, head of the Department of Neuropsychology and Psychopharmacology, and faculty of psychology and neuroscience, Maastricht University |
"AI is more than a buzzword — it’s transforming drug discovery and clinical research, driving real-world impact. As CScO at Astellas, I’m focused on how we leverage AI to accelerate drug discovery and development — for both small molecules and advanced modalities like cell and gene therapies — so that we can deliver high-quality assets into clinical development with greater speed and efficiency. “A great example of this would be ASP5502 – our STING inhibitor that is currently in Phase 1 clinical development for primary Sjögren's syndrome (PSS), a chronic autoimmune disorder causing symptoms such as dry eyes and mouth. AI technology helped Astellas shorten the usual two-year optimization process to just seven months, allowing us to begin clinical development more quickly. While we’re still in the early stages of realizing AI’s full potential, its role in accelerating the translation of science from the bench and into the clinic is undeniable.” |
|
“AI technology can transform how we approach clinical trials today by offering new ways to accelerate evidence generation. In our inflammatory bowel disease (IBD) research, machine learning provides the potential for more accurate and objective assessments of disease severity and treatment response, reducing statistical noise in evaluating patients’ eligibility and efficacy measurements. “Leveraging advanced AI algorithms from technology partners like Iterative Health, we can refine endoscopic assessments and enhance how results are interpreted — a key factor in clinical research. Harnessing the power of machine learning allows us to optimize patient selection, advance the efficiency and accuracy of clinical assessments, and improve treatment pathways. Our goal in using this approach is to strengthen clinical trial design and inform real-world decisions about therapy use and sequencing, ultimately driving progress toward precision medicine in IBD. |
Vijay Yajnik, MD, Ph.D., vice president, head of gastroenterology, U.S. medical affairs, Takeda |
“Driven by curiosity — and paired with a wonderful collaboration with Kelly Fitzgerald at WCG — I suggested the use of AI to ignite a project to benefit the industry writ large. I asked Kelly if WCG had data revealing what categories of payments sponsors were paying to participants and how that picture looked across therapeutic areas and phases of studies. I learned it’d be a manual process. I then asked Kelly if they could use AI and, boom, the project took off, leading to a joint co-authored publication in the Journal of Clinical Studies. ChatGPT4 was used to extract payment terms from 14,000 informed consent forms from 2019–2024 and then allocated to various payment categories. The average amount per study visit was calculated. “This, along with the metadata associated with the studies, allowed us to analyze trends in the data. This was a first-of-its-kind data study to understand how sponsors were paying participants over a five-year period. This is valued baseline data, revealing gaps and much-needed improvement by way of participant payments for the industry.” |
|
“At the Boston-based MIT Jameel Clinic, our researchers use AI to analyze mammograms to accurately predict patients’ risks of developing breast cancer within the next five years. This state-of-the-art AI model, Mirai, assigns a personalized risk score to each mammogram, helping clinicians determine when a patient should return for their next screening or when early intervention is needed. Mirai has been adopted by 44 hospitals around the world, with 1.8 million mammograms validated. Mirai is paving the way for breast cancer clinical research thanks to its high accuracy maintained across different patient demographics, a marked contrast to the more biased risk assessment approaches currently being used on patients. "The Jameel Clinic has also developed similar AI models to aid in risk prediction for other cancers, including pancreatic and lung. This technology is transforming the healthcare industry and has the ability to significantly reduce costly and invasive treatment options for patients.” |
|
“It's important to know that AI is really a broad topic that covers any technology that mimics human learning, decision-making, or performance, which can range from simple decision-making algorithms ("if this, then do that”) to much more complex systems. Even within that umbrella, there is further segmentation between things like machine learning and deep learning. There are many variations of both methodologies, and many are currently at work in clinical research. “For example, we are developing machine learning methods to recognize patterns in a patient’s anesthesiology monitoring during surgery to predict dangerous drops in blood pressure before they can occur, allowing an anesthesiologist to intervene and prevent serious adverse events to the patient. Another example is the use of deep learning for analyzing medical images. For example, it can automatically segment (a process called "contouring") anatomical regions such as the liver, pancreas, or a tumor within the image. Automation of the segmentation task using DL permits clinical researchers to analyze, for example, the tissue response to treatment. “Practical examples abound in both research and clinical use, and adoption at this still early stage of emerging technology varies among healthcare systems within the U.S. and globally.” |
|
“While there’s plenty of hype around AI, we’re seeing tangible applications in trials that enhance efficiency and accessibility. AI is applied at just about every stage in the clinical trial workstream. At inception, it serves as an essential tool in helping to develop protocols quickly. Algorithms are used to predict study outcomes based on historical clinical trial data, as well as to write protocols in line with applicable regulatory frameworks. AI also improves screening conversion rates and accelerates the recruitment process by analyzing EHRs to pinpoint patients who already meet key eligibility criteria. These are just a few examples of how AI is moving beyond the buzz to streamline research." |
|
More AI Advances In Clinical Research
Interested in learning more about AI in clinical research? Catch up with these guest experts who shared their insights on AI and other tech throughout 2024.