What Is Radiology–Pathology Correlation?

A radiologist looks at a study, describes a finding, and assigns a level of suspicion. Days or weeks later, a biopsy or resection produces the actual answer. Radiology–pathology correlation — often shortened to rad-path correlation — is the deliberate act of reviewing those two things side by side: the imaging impression and the pathology outcome.

It sounds simple, and conceptually it is. In practice, it is one of the most valuable — and most neglected — feedback loops in diagnostic medicine.

A working definition

Rad-path correlation compares the initial impression (what the imaging study reported and how confidently) with the ground-truth result (what the pathologist found in tissue). The comparison can be made at the level of a single case — "MRI called this PI-RADS 3; the biopsy showed Gleason 5+4 adenocarcinoma" — or aggregated across many cases to describe how a modality, protocol, individual radiologist, or the entire group is performing.

Why radiology–pathology correlation matters

Radiology is one of the few clinical disciplines where the practitioner routinely makes consequential decisions and then rarely learns how they turned out. The pathology result lands in the electronic health record, attached to a different encounter, and the radiologist who made the original call almost never sees it. Rad-path correlation exists to reverse that.

Diagnostic accuracy and quality assurance

Systematically comparing impressions with outcomes is the only way to measure diagnostic performance with real ground truth. It converts subjective confidence ("I think this is benign") into an evidence base — sensitivity, specificity, and predictive values that a department can actually act on. That is the backbone of a modern radiology quality-assurance program.

Education, peer learning, and CME

Every correlated case is a lesson. A concordant result reinforces a good habit; a discrepant one is a chance to recalibrate before the same mistake recurs. Because the feedback is tied to the physician's own prior work, it is far stickier than a lecture — and increasingly, this kind of practice-based learning can be structured to earn continuing medical education (CME) credit.

Patient safety

Discrepancies caught systematically can reveal patterns — a protocol that under-samples a region, a reporting habit that buries the actionable line, a radiologist who consistently over- or under-calls a particular entity. Surfacing those patterns early is a patient-safety win, not just an academic exercise.

Accreditation

Rad-path correlation is increasingly recognized in formal quality frameworks. The American College of Radiology (ACR) lists radiology–pathology correlation among the advanced criteria for its Diagnostic Imaging Center of Excellence (DICOE) designation. We cover that in detail in a dedicated guide (below).

Concordance vs. discrepancy

Every correlated case falls somewhere on a spectrum between agreement and disagreement:

• Concordance — the imaging impression matched the pathology. Reassuring, and useful for confirming that a reader or protocol is well calibrated.
• Discrepancy — the impression and the outcome diverged. These are the highest-yield cases: a benign-appearing lesion found to be malignant, or an aggressive-looking finding that proved indolent.

A mature program does not treat discrepancies as failures to be hidden. It treats them as the most instructive events available — provided they can be found reliably in the first place.

How rad-path correlation has traditionally been done — and why it breaks down

Historically, rad-path correlation has been manual and opportunistic. A radiologist remembers a case, hunts down the pathology report, and compares by hand. Tumor boards and multidisciplinary conferences correlate a curated handful of cases. Periodic audits pull a sample for review.

All of these are valuable, and all of them share the same limitation: they only capture the cases someone happens to go looking for. The reasons are structural, not a lack of will:

• The pathology result arrives days or weeks later, on a separate encounter, with no automatic link back to the imaging that prompted it.
• Finding the matching prior report means manual chart review — time most radiologists do not have.
• Without a systematic process, rad-path correlation skews toward memorable or interesting cases and misses the routine ones, where quiet, systematic error tends to hide.

The net effect: the overwhelming majority of the feedback that could exist never does.

Automating radiology–pathology correlation

The bottleneck is not judgment — radiologists are perfectly able to learn from a correlated case. The bottleneck is assembly: reliably pairing each pathology result with the prior imaging that described the same finding, across every case, without hours of manual chart review.

That is a natural fit for automation. Natural-language processing can read both radiology and pathology reports, extract the clinically meaningful concepts from each, and match a pathology result to the earlier imaging study that described the same lesion — then present the pair to the radiologist who made the original call. What was an ad-hoc, memory-dependent task becomes a systematic one that runs across an entire practice.

Keep going

This is the foundation. Three related guides go deeper on the parts practices ask about most: how rad-path correlation factors into ACR DICOE accreditation, how to build a correlation program without adding manual work, and using it for resident education and CME.

Frequently asked questions