Radiology companies explained: the five types compared

Radiology companies are not one industry. The term covers a medical practice employing physicians, a private-equity platform managing hundreds of such practices, a remote reading service, a chain that owns scanners and buildings, and a software vendor whose product has never signed a report. A director shopping for coverage gets quotes from three of these types in one week, priced in units that do not compare.

This page is the map: the five types, what each actually sells, who reads the studies in each model, how the money flows, and which type answers which problem. One declared interest up front: AstraRad is the third type, a subspecialist teleradiology practice, and the map notes where that model fits and where a different type serves you better. If your problem is already defined as unread studies, a written rate card answers it within one business day of a request.

What counts as a radiology company? The five types

Type	What it sells	Examples	Who employs the reading radiologist
Radiology practice group	Professional interpretation, on-site presence, hospital contracts	Thousands of local and regional groups	The group
Practice-management consolidator	Ownership or management of many groups at scale	Radiology Partners, US Radiology Specialists	The owned or managed practice
Teleradiology company	Remote interpretation as a service, per study	vRad, AstraRad, StatRad, ONRAD	The teleradiology practice
Imaging center operator	Scanning capacity: facilities, equipment, technologists	RadNet, SimonMed, Akumin	Varies: employed, contracted, or outsourced
AI and software vendor	Algorithms and workflow tools that assist reads	Aidoc, Viz.ai, Rad AI	Nobody: no AI signs a report

Two adjacent types complete the picture, and both are easy to misfile. Mobile imaging operators drive portable X-ray and ultrasound units to nursing facilities, correctional sites, and homebound patients; they own trucks and technologists, and nearly all of them buy their interpretation layer from type three, which makes them teleradiology's most geographically complicated customers. Staffing firms place locum radiologists into physical vacancies by the day or week; they sell a person on site, which no remote service replicates, at day rates that make them the expensive option for any gap measured in months. The head-to-head is worked through in the locum vs teleradiology analysis.

The scale behind the whole map is a physician workforce that consolidation cannot expand: a 2026 study in the Journal of the American College of Radiology counted 38,306 practicing radiologists as of 2022, per the pipeline analysis on PubMed, while imaging volume grows faster than residency slots. The ACR's workforce update treats the shortage as structural, and every company type on this page is, in one way or another, a machine for allocating that scarce reading capacity. Which is why the reader question comes first.

Who actually reads your studies?

Strip away the corporate layer and every study still ends with one credentialed physician signing one report. The types differ in where that physician sits and who pays them, and the differences decide your economics and your quality levers.

In a practice group, the group's radiologists read, on site or remotely, and the group typically bills payers for the professional fee under its hospital or facility contract. Your lever is the contract and the relationship; your risk concentrates in the group's own staffing, which is exactly what fails when a group gives notice.

In a consolidator, the same reading happens inside an owned practice, with corporate scale behind scheduling, recruiting, and technology. The entity structure deserves attention during contracting: the parent company and the practice that provides the reads are often separate legal entities, so confirm which practice entity is on your agreement.

In a teleradiology company, readers are employed or paneled by the service, licensed in your patients' states, credentialed at your facility, and the facility is invoiced per report while your side bills the payer. The whole operational anatomy, worklist engines, licensing departments, QA lines, is laid out in inside a teleradiology company. Report tier is the split that matters most inside this type: preliminary reads leave your own physicians re-reading in the morning, while AstraRad's model is final signed reports from fellowship-trained subspecialists at every hour, with STAT under 4 hours by contract and 2-hour or 1-hour STAT available on request at a higher multiplier, per the SLA.

In an imaging center operator, interpretation arrangements vary by market: employed readers, a contracted local group, or outsourced overflow. RadNet, which reports 442 centers as of mid-2026, is one large operator; smaller operators frequently buy the reading layer from types one and three.

In an AI vendor, nobody reads. Cleared algorithms flag, prioritize, measure, and draft under an interpreting physician's signature, and the FDA's AI-enabled device list now carries over 1,100 radiology clearances, nearly all assistive. The vendor landscape and its buying questions fill the AI radiology companies guide.

How the money flows through each type

Follow the professional fee and the five types stop looking interchangeable, because each one monetizes a different asset.

Practice groups and consolidators bill payers. The group holds the facility contract, reads the study, and files for the professional component of each CPT code; the hospital pays little or nothing directly, and sometimes receives coverage stipends in the other direction as call coverage grows scarce. The group's asset is the payer relationship plus the exclusive contract, and rising stipend demands are usually the first visible symptom of the workforce math tightening underneath it.

Teleradiology companies invoice the facility. In the common model the facility pays per signed report and its own revenue cycle bills payers for the professional component, which keeps the vendor out of your payer relationships entirely. The vendor's asset is reading capacity: licensure, credentialing, and scheduled subspecialists. This is why per-report pricing exists at all, and why comparing a teleradiology invoice against a group's stipend requires converting both to cost per study; the arithmetic lives in the teleradiology cost guide.

Imaging center operators bill globally. Owning the scanner means owning the technical component, the larger share of most outpatient reimbursement, with the professional component handled by whoever reads. Their asset is the equipment, the real estate, and the referral flow.

AI vendors invoice subscriptions. With narrow exceptions, no payer reimburses the algorithm, so the spend is operational and the vendor's asset is the clearance plus the integration. The exceptions, an add-on payment here, a dedicated CPT code there, are notable precisely because they are rare.

A buyer comparing across types is therefore comparing different units: a stipend, a per-report rate, a joint-venture distribution, a license fee. Normalize everything to annual cost against your own study counts before letting any two quotes share a spreadsheet, and be suspicious of any proposal whose unit resists the conversion.

How are radiology companies consolidating?

Ownership across every type has changed enough in three years that a 2023 vendor map is out of date. Announced moves a buyer should know: Radiology Partners acquired vRad from MEDNAX in 2020. I-MED Radiology Network announced its acquisition of StatRad in July 2024, in a release describing the deal as its entry into the US market. ONRAD states that it acquired Direct Radiology in December 2024. And AI consolidation runs through platforms: single-algorithm companies increasingly sell through or into the marketplaces of larger vendors.

None of this is automatically bad for buyers, and some scale genuinely helps: deeper night benches, better recruiting, real engineering. The contracting consequences are what deserve attention, and they are the same three every time. Know which legal entity signs your agreement and which parent stands behind it. Negotiate a change-of-control termination right, because the vendor you selected can become a different company mid-term without your consent. And re-run reference checks after any acquisition touching your account, since service terms can change under a new parent. What the major teleradiology names state publicly is listed in the company overview.

Independence is the other side of the graph, and worth naming as a fact rather than a virtue: AstraRad is an independent practice, unowned by any consolidator, roll-up, or platform, which means the entity on the contract is the entity doing the work. Whether that matters is your call to make; that it is verifiable is the point.

Which type of radiology company do you need?

Start from the failure, because each type exists to absorb a different one.

Your problem	The type that solves it	Where to go deeper
Nights, weekends, subspecialty gaps, overflow	Teleradiology company	Best pick by facility type
Whole-department staffing with on-site presence	Practice group or consolidator	In-house vs outsourced analysis
A defined absence with a return date	Locum staffing	Locum vs teleradiology
Scanner capacity, outpatient expansion	Imaging center operator	Market-specific; the operators quote directly
Deep queues, triage attention, reporting speed	AI vendor, layered on human reads	The AI vendor landscape

Mixed problems take mixed answers, and the most common modern configuration is exactly that: a local group or employed core for daytime and procedures, a teleradiology practice for nights, subspecialty, and overflow, and possibly an AI layer on the facility's own worklist. The types interlock more than they compete, which is why evaluating each on its own axis beats forcing one vendor to be everything. Cross-type quotes should still be normalized to one number, cost per signed study on your own volume, and checked for what US-based, coverage, and licensure claims actually mean; the US teleradiology companies guide covers that verification for type three.

Three common misfilings, and what they cost

The map earns its keep by preventing category errors, and three recur constantly. A hospital asks an AI vendor to fix an unread overnight list; the algorithm reorders a queue that still has no reader at the end of it, and six months of license fees later the list is unchanged. An imaging center asks a locum firm to cover a permanent subspecialty gap; the day rates that make sense for a four-week absence quietly become the department's largest line item by month five. A facility asks its teleradiology vendor for on-site fluoroscopy supervision; no remote practice can stand in that room, and the request signals the facility actually needs a local group relationship alongside its remote coverage.

Each misfiling shares a root: the buyer priced vendors within one type instead of first checking whether the type itself matched the problem. Ten minutes with the table above is cheaper than any of the three lessons, and every one of them was avoidable at the RFP stage with a single question about what the vendor cannot do.

Reading the map before you buy

The taxonomy compresses to three questions that sort any radiology company you meet. Who employs the physician who will sign my reports? What exactly is invoiced, per study, per shift, per subscription, or per algorithm? And which entity, under which owner, is accountable when the service misses? Ask them of a practice group, a consolidator, a teleradiology service, an imaging operator, and an AI vendor, and you will get five different answers, which is the whole point: the answers, side by side, are the market map for your specific facility.

For type three, AstraRad's answers are standing: our own subspecialist panel signs every report, the invoice is one line per signed report with no minimums and no platform fees per the pricing model, and the accountable entity is the independent practice itself. A written rate card against your case mix arrives within one business day of a request, which is a reasonable way to put one concrete number on the map before talking to the other four types.