Why Clinical Trial Imaging Needed Infrastructure Before It Needed AI
When Forbes, Fierce, or their peers publish annual lists of the world’s most innovative companies, the winners tend to share something in common.
They’re visible.
A breakthrough AI model. A revolutionary medical device. A life-saving therapy. Innovation, at least on the surface, is often associated with the things we can see.
But history suggests something different. Some of the most important innovations ever created were almost completely invisible. Few people celebrate databases, payment networks, or the operational systems that make complex industries work. Yet modern society depends on them. Infrastructure rarely gets recognized in the moment. It gets recognized years later, when we realize how much progress depended on it. The rarely spotlighted world of clinical trial imaging is experiencing a similar moment.
The Problem Nobody Wanted to Solve
For decades, imaging has played a critical role in clinical research. It’s pervasive across therapeutic areas, and nowhere more so than in oncology, where the efficacy endpoints regulators rely on, like objective response rate and progression-free survival, are derived from imaging through standardized criteria such as RECIST.1,2 Sponsors spend significant resources collecting imaging data to evaluate efficacy, monitor safety, and support research and regulatory objectives.
The images themselves have become increasingly sophisticated, and the technology surrounding image acquisition has advanced dramatically. We also regularly see artificial intelligence applied to image analysis, to the point that AI in imaging no longer feels novel.
But behind the scenes, one critical problem remained largely unchanged. The operational infrastructure managing imaging workflows never evolved at the same pace. Across the industry, trial imaging often depended on disconnected systems, spreadsheets, email chains, manual reconciliation, institutional knowledge, and countless handoffs between sites, CROs, sponsors, and core labs.
Everyone knew the complexity existed and suffered through the consequences, yet the industry largely accepted the problem as unavoidable. The result was a hidden operational burden that introduced delays, increased costs, and created significant data quality risks.
The scale of the problem is easy to underestimate. A peer-reviewed analysis of RECIST 1.1 reads looked at the discordance rate between the results from the local treating investigator and a central reviewer, with 36.7% of all analyses having disagreement and requiring an adjudicator to break the tie. The causes of these discrepancies were, among other things, reviewers often working from different baseline scans or selecting different target lesions.3 Audits point in the same direction: a substantial share of results, often a quarter to almost half, carry errors, missing information, or workflow exceptions that require manual intervention before the data can be trusted.4 Downstream, those errors create unnecessary costs, delays, and risks for sponsors.
What’s striking about those numbers is that they are not the result of careless teams or inadequate expertise. In most cases, highly skilled professionals are working incredibly hard to keep studies moving forward. The problem is that they are relying on systems and processes that were never designed to manage the scale and complexity of modern research and imaging operations.
Clinical Trial Imaging Had Tools. It Never Had Infrastructure.
Historically, organizations approached trial operations the same way many industries did before infrastructure emerged: they solved problems one at a time. When image transfers became difficult, a process was put in place to manage them. When reader management became more complex, another system was added. As reconciliation challenges surfaced, spreadsheets multiplied. Reporting requirements led to more reports, more trackers, and more manual oversight.
Each solution addressed a real need, and each process made sense in isolation. But over time, organizations found themselves managing an ever-growing collection of disconnected workflows rather than a cohesive operational system. It’s the equivalent of building a city by continuously adding roads wherever traffic appears. Eventually, you have plenty of roads, but no thoughtful network. Movement becomes harder, not easier.
Clinical trial imaging reached a similar point. The individual components existed, but there was no coordination layer to orchestrate them, connect the data, or relieve the operational burden. As studies grew larger and more complex, that missing layer became increasingly difficult to ignore.
The Missing Layer
Every industry eventually outgrows its patchwork solutions. Banking couldn’t scale without payment networks. Healthcare needed electronic medical records. Clinical research needed electronic data capture. These weren’t flashy innovations when they first appeared. In fact, many seemed almost mundane compared to the technologies they supported. What made them transformative wasn’t what they did individually. It was what they enabled collectively.
Clinical trial imaging never developed a comparable workflow and data management layer. For years, the industry continued to operate through a combination of people, manual processes, and siloed knowledge. That worked until it didn’t.
Good infrastructure quietly removes friction. It creates consistency where there was once variability. It turns fragmented information into trusted data, allowing experts to spend less time managing operational complexity and more time applying their expertise.
Making Errors Visible
One of the most revealing moments for organizations implementing a modern imaging workflow platform is not what they gain, but what they discover. As workflows become standardized and data becomes visible across the imaging ecosystem, longstanding operational issues that were previously hidden begin to surface: mismatched records, missing information, protocol deviations buried in manual processes, and bottlenecks that teams sense but cannot quantify.
For years, these issues existed in plain sight across emails, spreadsheets, shared drives, and disconnected systems, but because they were fragmented, few organizations could see the full picture. Once those workflows are connected, the picture becomes impossible to ignore. For many organizations, the findings are surprising. Error rates assumed to be isolated incidents turn out to be systemic, surfacing across a meaningful share of the imaging records in a study. At three NCI-Designated Comprehensive Cancer Centers, that discovery process was published: site read error rates as high as 50% dropped below 2% once imaging workflows were standardized and connected, and re-review rates fell from 30% to 2%.
There is an irony worth naming here. The same infrastructure built to operate quietly in the background is what first drags these hidden problems into the light. The discovery can be uncomfortable. It is also the entire point. Problems that stay invisible cannot be fixed, while problems that can be measured can be solved.
The Future Belongs to the Invisible
The clinical research industry is in an era of growing complexity. Studies are expanding across more sites, countries, and patient populations, while protocols continue to grow in sophistication. Imaging endpoints are becoming increasingly important across therapeutic areas, from oncology to dentistry. At the same time, artificial intelligence is creating entirely new opportunities for analysis and decision-making.
But AI, analytics, and automation all depend on the same foundation: reliable, standardized, traceable data. Without that foundation, advanced technology can amplify workflow problems rather than solve them. Every major technology transformation eventually reaches a point where it can no longer scale through people, spreadsheets, and disconnected systems alone. At that point, a new operational layer emerges, one that standardizes workflows, connects data, increases visibility, and enables new levels of efficiency and collaboration.
Clinical trial imaging is reaching that point now. The emergence of Yunu’s dedicated workflow and data management platform signals that the field is finally realizing the layer it has long lacked. With adoption that indicates real market traction, the built-for-purpose platform is quietly and quickly becoming the foundation on which research imaging is experiencing this needed transformation. The intent is straightforward: connect sites, readers, sponsors, and CROs within a single operational environment rather than across a patchwork of spreadsheets and inboxes. The value of that layer does not show up in flashy demonstrations or announcements. It shows up as a standard for cleaner data, more access to experts, fewer errors, faster study execution, stronger oversight, and the ability to catch operational issues before they become costly problems.
More than a third of the NCI-Designated Comprehensive Cancer Centers in the United States have already moved their imaging trial operations onto Yunu.5 Adoption at that scale is one of the clearest signals yet that the field was ready for this shift. Not because infrastructure is exciting, but because dependable and thoughtful operations are essential for success.
History rarely celebrates infrastructure when it is first built. It celebrates what becomes possible afterward. The next era of clinical research will not be defined solely by smarter algorithms or more advanced imaging technologies. It will be defined by the operational foundation that allows those innovations to scale.
That is the paradox of transformative infrastructure: when it works well, it almost disappears. Yet once it exists, it becomes impossible to imagine operating without it. That may be the clearest sign of an innovation that changes an industry.
References
1. Ruchalski, K., Braschi-Amirfarzan, M., Douek, M., Sai, V., Gutierrez, A., Dewan, R., & Goldin, J. (2021). A primer on RECIST 1.1 for oncologic imaging in clinical drug trials. Radiology: Imaging Cancer, 3(3), e210008. https://doi.org/10.1148/rycan.2021210008
2. U.S. Food and Drug Administration. (2018). Clinical trial endpoints for the approval of cancer drugs and biologics: Guidance for industry. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-trial-endpoints-approval-cancer-drugs-and-biologics
3. Beaumont, H., Evans, T. L., Klifa, C., Guermazi, A., Hong, S. R., Chadjaa, M., & Monostori, Z. (2018). Discrepancies of assessments in a RECIST 1.1 phase II clinical trial—Association between adjudication rate and variability in images and tumors selection. Cancer Imaging, 18, Article 45. https://doi.org/10.1186/s40644-018-0186-0
4. Yunu. (2024). 50% imaging error rate in clinical trials discussed by expert panelists from 5 NCI-designated cancer centers. https://www.yunu.io/blogs/post/50-imaging-error-rate-in-clinical-trials-discussed-by-expert-panelists-from-5-nci-designated-cancers
5. Yunu. (2024). Yunu emerges as industry leader to address the way imaging data impacts breakthrough therapy development. https://www.yunu.io/blogs/post/yunu-emerges-as-industry-leader-to-address-the-way-imaging-data-impacts-breakthrough-therapy-develop
