A patient comes in for a CT pulmonary angiogram for suspected pulmonary embolism. The CTPA is negative for PE. But on the same series, there's a 7mm solid pulmonary nodule in the right lower lobe, a 1.8cm adrenal nodule on the right, and a small amount of free fluid in the pelvis. The ordering ED physician gets a report that says: no PE identified, with the incidental findings listed in the body of the report.
What happens next? In most departments, the honest answer is: it depends on whether anyone reads the report carefully. There's no structured follow-up trigger, no automatic flag to the patient's PCP, no tracking system to verify the 3-month CT recommended in the report's final line was actually ordered. The radiologist has done their job. The follow-up gap is a system problem, and it's a widespread one.
How Prevalent Are Incidentalomas, Really?
The term "incidentaloma" was coined to describe clinically significant findings discovered incidentally on imaging obtained for an unrelated indication. The most studied categories are adrenal incidentalomas, pulmonary nodules, thyroid nodules, renal masses, and ovarian lesions — but incidental findings on CT and MRI span virtually every organ system.
Published prevalence estimates vary considerably depending on the study population and imaging protocol, but the numbers are not trivial. In adults over 50 undergoing abdominal CT for any reason, adrenal incidentalomas are found in roughly 3-5% of studies. Pulmonary nodules appear on approximately 30-35% of chest CT studies in older adult populations with significant smoking histories — with the vast majority being benign, but requiring a follow-up protocol to establish that. Renal masses are found incidentally on roughly 10% of abdominal CT studies and require characterization to differentiate simple cysts from potentially malignant lesions.
A busy imaging center processing 150 CT studies per day will generate several dozen incidental findings requiring follow-up every week. Most of these findings are benign — the adrenal nodule is a lipid-rich adenoma, the pulmonary nodule is a calcified granuloma, the renal mass is a Bosniak I simple cyst. But the process of determining that requires either a single definitive characterization study or a follow-up interval comparison, and that process has to be tracked somewhere.
Where the Workflow Gap Actually Lives
The incidental findings problem isn't primarily a detection problem. For common incidentalomas in adult populations, radiologists are reasonably good at identifying and reporting them. The problem is downstream: connecting the finding in the radiology report to a follow-up action that actually happens.
The workflow gap has three components:
Communication handoff. Radiology reports enter the EHR and become the ordering clinician's responsibility. In a busy ED or outpatient setting, a 15-item CT report has a lot of competition for the ordering physician's attention. The three-line incidental finding paragraph at the end of a report focused on the acute indication gets read by some clinicians and not others. This isn't negligence — it's triage.
Recommendation clarity. Radiology reports vary considerably in how specifically they phrase follow-up recommendations. "Follow-up in 3 months" is less actionable than "3-month low-dose CT of the chest per Fleischner Society 2017 guidelines for a 7mm solid nodule in a high-risk patient." The latter gives the ordering clinician a specific study to order and a specific reason. The former requires them to initiate a follow-up workup with minimal guidance on what that means.
Tracking and closure verification. Even when the recommendation is clear and the ordering clinician intends to act on it, the completion of follow-up imaging isn't tracked at the department level. If a patient doesn't return for the 3-month follow-up CT, most radiology departments have no mechanism to detect that and close the loop.
Society Guidelines Exist — Adoption Is Inconsistent
The major radiology societies have published structured management guidelines for the most common incidental finding categories. The Fleischner Society guidelines for pulmonary nodules (most recently updated in 2017 for solid nodules and 2017 separately for subsolid nodules) provide size- and morphology-based follow-up recommendations stratified by patient risk factors. The ACR Incidental Findings Committee has published recommendations for adrenal, thyroid, liver, and other organ findings. LI-RADS provides a structured categorization system for liver observations in patients at risk for hepatocellular carcinoma.
The existence of these guidelines doesn't translate automatically into consistent application. In a department without structured reporting templates, whether a radiologist applies Fleischner criteria versus a more conservative institution-specific protocol versus an informal personal practice pattern depends on training and habit. When a department reads across multiple radiologists and locum tenens coverage, that variability compounds.
Structured reporting for incidental findings — where the DICOM viewer or reporting system prompts the radiologist to enter a finding category, size, and guideline-based recommendation — reduces this variability. But structured reporting alone doesn't solve the downstream tracking problem. The report is still just a document that relies on human follow-through.
What a Functional Incidentaloma Workflow Looks Like
The departments that handle incidental findings well tend to have three things in place that most departments don't:
A dedicated registry or tracking system. This is sometimes a module within the hospital's EHR, sometimes a standalone radiology-operated registry, and sometimes a manually maintained database. The key requirement is that when a radiology report contains an incidental finding with a follow-up recommendation, an entry is created in the tracking system with the finding type, recommended follow-up study, recommended timeframe, and responsible ordering provider. Someone owns that entry until the follow-up is completed and the finding is resolved or classified.
Explicit report language that triggers a follow-up action. Structured radiology reporting templates can tag specific phrases or data fields that automatically populate the tracking system when a report is signed. This removes the manual transcription step that creates leakage between the report and the registry. It requires either EHR-native structured reporting or an interface between the radiology reporting system and the tracking database.
A closure process. Someone — either in radiology or in primary care coordination — reviews open tracking entries at regular intervals and contacts patients or ordering providers when recommended follow-up hasn't been completed within the specified window. In some institutions this is a radiology nurse coordinator role. In others it's a joint function between radiology and care management. The specific model matters less than the existence of a closure loop with clear ownership.
We're not saying this is easy to implement — it requires administrative infrastructure that many departments don't have and EHR configuration work that can take months. But the alternative — a large, unmeasured leakage rate between incidental finding and completed follow-up — carries real medicolegal and patient safety implications. Several high-profile radiology malpractice cases in the last decade have involved lung nodules or renal masses that were reported incidentally and then never followed up, with the finding subsequently representing a stage-advancing delay in cancer diagnosis.
Where AI Pre-Annotation Fits In
The connection to pre-annotation tools like Neurmorph is indirect but real. The primary value of surfacing a 7mm pulmonary nodule with a pre-populated size measurement and Fleischner risk stratification isn't just that the radiologist finds it faster — it's that the annotation includes the structured data elements that a follow-up tracking system needs. A size, a morphology classification, a guideline category, and a recommended follow-up interval can flow from the annotation directly into a report template and from there into a tracking registry, rather than existing as free text that requires parsing.
This doesn't solve the organizational problem of whether the tracking registry exists and whether someone is closing the loop. It does reduce the friction in the data capture step. For departments that are building toward a structured incidentaloma management program, that friction reduction matters: the harder it is to generate structured finding data, the less consistently it gets done.
The incidental findings problem is ultimately about communication and process ownership more than detection. The scan produces the finding; the report communicates it; the system tracks it; the follow-up closes it. Every link in that chain needs to function for the finding to result in the right patient outcome. Detection is the link that gets the most attention because it's the most technically interesting. The others are where most of the actual work remains.