Medical Imaging
3
min read
Beyond Image Collection: What Central Imaging Really Delivers in Clinical Research
A central imaging core lab delivers far more than image collection: imaging charter and protocol design, standardized acquisition, independent blinded reads and adjudication, quality control, data management and submission-ready documentation. The real deliverable is a defensible imaging endpoint, reliable enough to support regulatory decisions, not a repository of scans.
Ask a procurement team what an imaging core lab does and the answer usually involves receiving, storing and reading images. Priced that way, one provider looks much like another, and the decision comes down to cost per scan.
That framing describes about a fifth of the work, and not the fifth that determines whether the endpoint holds up.
The Misconception: Central Imaging = Image Logistics
Logistics is real work. Images have to be de-identified, transferred, cataloged and archived, and doing that badly causes genuine problems. It is also the part of the service that is easiest to specify and easiest to replicate.
The output a sponsor needs is different: a dataset in which every response assessment was produced by a documented process, applied identically to every patient, and reconciled with the clinical database at lock. Images are the raw material. The endpoint is the product.
The distinction becomes concrete when something goes wrong. A logistics provider can tell you a scan arrived. A core lab can tell you the scan was acquired against protocol settings, passed quality control, was read by two qualified readers under the charter methodology, produced discordant assessments adjudicated by a third, and that the classification is traceable to the source images.
What an Imaging Core Lab Really Delivers
Imaging charter and protocol design
The work starts before the protocol is final. FDA guidance on clinical trial imaging endpoint process standards expects the imaging charter to detail the trial's imaging methodology and to describe how sources of imaging bias and variability are controlled. Writing it forces decisions the protocol often leaves implicit: which response criteria apply, what equipment settings each site must use, which deviations are acceptable, how patients are prepared and positioned, how wide the timepoint windows are, and what the read design will be.
A core lab involved at this stage prevents problems. Involved after enrollment starts, it can only document them.
Independent reads and adjudication
Reading is where the endpoint is created. The deliverable is not a radiologist's opinion but a classification produced under a fixed methodology by readers independent of the sites, blinded to treatment assignment, trained on the criteria in the protocol and qualified against prespecified thresholds.
Where the read design uses two independent readers, discordance is resolved by an adjudicator under rules set in the charter. The adjudication rate becomes a quality indicator in itself: stable over the study, it suggests the criteria are being applied consistently; rising, it usually points to reader drift or an ambiguity in the charter.
QC and data management
Quality control runs continuously, not at the end. Incoming images are checked against charter criteria, queries are raised while the visit window is still open, deviations are logged. The guidance also recommends periodic monitoring of acquisition, storage and transfer, and periodic blinded re-testing of readers using predefined image sets interspersed with trial images.
Data management covers de-identification integrity, DICOM handling, query resolution, reconciliation with the clinical database and archiving. Images are expected to be retained as a component of subject care and, per the guidance, for no less than two years following approval.
Regulatory and submission support
The final deliverable is documentation as much as data: the charter and its amendments, site and reader qualification records, training records, quality control logs, the audit trail from source image to final classification, and datasets structured for submission.
Invisible while a study runs, decisive when a question arrives during review. A reviewer asking how a progression date was determined should receive a documented chain, not a reconstruction.
The Value Chain, End to End
Phase | What the core lab does | What the sponsor gets |
|---|---|---|
Protocol design | Defines imaging methodology, response criteria, read design and acquisition standards | An imaging charter, and a protocol whose imaging section is executable |
Study start-up | Qualifies and trains sites, configures transfer, qualifies readers | Sites able to produce conforming images from the first patient |
Conduct | Receives, de-identifies and quality-controls images; raises queries; runs blinded reads and adjudication | Endpoint data produced continuously, with deviations caught inside the visit window |
Ongoing quality | Monitors reader performance, runs consensus sessions and blinded re-testing, retrains where needed | Consistency maintained across a long enrollment period |
Database lock | Reconciles imaging data with the clinical database, locks and exports | A dataset that matches, and a documented trail behind every value |
Submission | Supplies charter, qualification records, QC logs and structured datasets | Answers to reviewer questions that do not require reconstruction |
What to Look for in an Imaging Partner
Six questions separate providers more effectively than a capability matrix.
Does the imaging team engage before the protocol is final? If the first conversation is about image transfer, the charter is being written around decisions that were already made.
Do the readers match the criteria? A protocol using PERCIST or RANO needs readers trained on PERCIST or RANO. Ask which criteria the panel actually reads against, and how many readers per criterion.
How is reader consistency measured during the study, not before it? Ask for the mechanism: performance monitoring, consensus sessions, blinded re-testing, retraining triggers.
What is the quality control turnaround? A query raised three months after acquisition is a record. A query raised in three days is a chance to re-scan.
Is submission documentation part of the scope or a separate project? Assembling qualification and QC records retrospectively is avoidable work.
Who owns reconciliation with the clinical database? Unclear ownership at this handover is one of the more common causes of a delayed lock.
Banook answers those questions with a specific operating model. Every study runs on three documents: an imaging manual covering site qualification, image acquisition and quality control; an imaging review charter covering the independent central read procedure, reader management, reading methodology and schedule; and, for radioligand programs, dosimetry documentation covering calibration, acquisition requirements and data analysis.
The reader panel exceeds 500 board-certified specialists worldwide, each with at least ten years of experience across radiology, oncology, neuroradiology, nuclear medicine, dermatology and cardiology, and consistency is managed through statistical performance monitoring, consensus sessions, retraining and phantom-based bias checks.
Modalities cover CT, MRI, PET, SPECT, X-ray, ultrasound, angiography, colonoscopy, cardiac MRI and DXA, across oncology, cardiology, neurology, nephrology, metabolic disease and orthopedics.
The group has supported more than 350 imaging studies, including pivotal trials, from early phase through Phase IV, and runs the imaging workflow on a platform developed in house with de-identification, automated and manual quality control, centralized query management and audit trails.
Review the full central imaging offering, or bring an imaging endpoint to the team while the protocol is still open. The strategic case is covered in Why Central Imaging Has Become a Strategic Asset in Clinical Trial Execution.