Why the First Scan Matters Most: IAG’s Romiesa Hagoug on Keeping Imaging Data Honest Across Multi-Site Trials

At Image Analysis Group (IAG), every image in a clinical trial moves through DYNAMIKA™, a single proprietary platform that handles image collection, quality control, central reading and reporting, with a full audit trail from upload to final results.

The gap between a site that follows an imaging protocol and one that thinks it does is not always visible in the images themselves, at least not at first. It shows up in DICOM headers, in IQC trackers, in the quiet trending of parameters that no single reviewer would flag as a problem. For sponsors running global trials with dozens of imaging centers, that gap is where data quality erodes.

Romiesa Hagoug has spent her career at the intersection of that problem. As Head of Project Management at IAG, she oversees the systems and processes that move imaging data from site acquisition through central review, quality monitoring, and final reporting. In this conversation, she walks through what a first scan actually tells you about a site, how DEXA calibration and MRI hardware changes are managed in real studies, what reader panel drift looks like when you catch it early, and what sponsors should be asking imaging vendors before they sign.

Moe: Why is the first scan a site submits so important to your team?

Romiesa Hagoug: The first scan at every site is potentially the most important. It really determines the effectiveness of the training we have provided, and whether or not the site has understood what is expected of them for that clinical trial and for accurate imaging.

That first image we receive is expedited for review. Our imaging specialists look at anatomy coverage, compliance with the clinical protocol, and whether it meets the objectives of the study. We review it in a lot of detail. It effectively becomes the qualification scan for that site. And then we provide feedback on how to improve going forward.

Moe: Why do sites most commonly fail that first submission?

Romiesa: It definitely varies depending on the type of images required and the therapeutic area. In more standard modalities, the issues tend to be administrative: a site forgot to upload supplementary information, or missed a sequence, or couldn't upload it correctly.

When you move into more advanced image analysis, you start to see issues with the sequences themselves. Maybe they don't follow the guidelines. Maybe the site has reverted to their usual processes rather than following the study-specific requirements. That is exactly why we think the first image is so important for that expedited review. We want to catch those issues early and go back to the site to correct them before they become a pattern.

The first scan at every site is potentially the most important. It really determines the effectiveness of the training we have provided.

Moe: How do you manage DEXA calibration variability across a live multi-site study?

Romiesa: Early on, we identify the scanner that will be used for the study and collect all the relevant information. We have a detailed site qualification process where we review the IQC trackers from the site and their standard procedures for monitoring, acquiring data, and managing quality control. That is a very integral part of our feasibility work.

When we qualify a site, we confirm that images are of suitable quality and that the site is following the IQC requirements. Throughout the study, we collect monthly IQCs to monitor for any changes to the scanner or any variability that could affect image quality and the standardization of the final analysis.

If a scanner does change during a study, we have defined processes for that. We would apply correction coefficients to standardize the data output from the old scanner to the new one. And for multi-center trials spanning multiple countries, a single cross-calibration phantom is shipped across all sites. That assists us in making sure the data remains homogeneous for the client's analysis.

Moe: How do you handle a site that upgrades its MRI hardware partway through a longitudinal study?

Romiesa: It is very much dependent on the therapeutic area and the type of analysis planned. For some analyses, particularly semi-quantitative or visual assessments, minor changes in hardware or software may not considerably affect image quality or the central analysis.

However, there are outputs that require the scanner not to change at all. Significant hardware or software upgrades can affect the image outcome and also our ability to analyze the data centrally.

These are all things we review with sites upfront. We understand the duration of the study, and we ask sites to commit to flagging any planned software or hardware changes in advance. That gives us a more proactive approach to those types of issues. Consistency and standardization across all sites is the most important thing.

Moe: How do you keep a global reader panel calibrated?

Romiesa: Depending on the study, the type of read, and the imaging criteria, there are various training and harmonization sessions employed. For certain read criteria, we use test cases. Readers go through those test cases, and we make sure they are reading within a certain range. Those cases are reviewed by our medical lead to limit variability and ensure readers are reading to a consistent standard.

We also have a detailed quality monitoring process within our studies. We review initial cases and then schedule reviews throughout the study to make sure readers continue to read consistently and accurately, with that medical oversight in place.

Moe: How did a real case of imaging data drift get caught in time?

Romiesa: We had a large phase two study with multiple modalities being acquired. The site had been trained and qualified. We then identified certain parameter changes in the MRI. They were not significant, and they did not affect image quality, but they were a trending matter. We were seeing this consistently across different patients at different time points.

On review, we realized it was one member of personnel who had not been trained on the study and had been acquiring images across several patients. Our proactive approach during routine quality control identified that an untrained person was acquiring images. We caught it early, identified that there was a lack of oversight at the imaging facility, and raised it to the clinical facility and the sponsor to correct it. Thankfully, it did not affect image quality severely, but we were able to catch it.

Moe: How does real-time quality monitoring translate into your day-to-day systems?

Romiesa: I think having real-time quality monitoring and oversight from an imaging perspective is the only way to make sure you have high-quality data that is suitable for analysis. Our imaging specialists, who oversee the process, are all radiographers by trade and by background. That is super important for making sure images are acquired consistently and correctly, to the needs of a radiologist and to the needs of central reading.

That is the only way to make sure your data is not just suitable now for what you need it to be, but also reviewable later on. Consistency and quality of data are paramount.

A key part of our DYNAMIKA™ platform is incorporating technical edit checks within the data. For example, images being uploaded are checked against the DICOM header from the site's qualification scan. That allows us, in real time and efficiently, to flag any small changes that may not affect image quality but could be worth raising as a trending matter. When it comes to reading, we build all the read criteria within DYNAMIKA™, with embedded edit checks to make sure readers are reading consistently and within a predefined range.

Having real-time quality monitoring and oversight from an imaging perspective is the only way to make sure you have high-quality data that is suitable for analysis.

Moe: Why should a sponsor ask an imaging provider about their systems before signing?

Romiesa: I think a very important question is what systems are being used. Different systems for image collection, queries, and analysis can each introduce errors, missing data, or mismatches. Within IAG, we mitigate that by having everything in one system. You have complete oversight of data and a full audit trail from the moment images come in to the moment reports are released.

I think it is a very significant thing for sponsors to make sure they fully understand how data travels through an organization before they receive their results.

Romiesa Hagoug is Head of Project Management at Image Analysis Group (IAG), where she oversees imaging operations across multi-site clinical trials.

Image Analysis Group (IAG) is an imaging CRO supporting sponsors across global, multi-site clinical trials, from site qualification through central review and database lock, powered by its DYNAMIKA™ platform. Learn more at www.ia-grp.com.