AI Agents for Radiology Practices: Automating Billing From Professional-Technical Component Splits to Interventional Procedures and Contrast Administration in 2026

Radiology is the highest-volume billing specialty in medicine — and the one where a single modifier error multiplied across thousands of studies per month creates six-figure revenue leakage that nobody notices. A 12-radiologist group reading 150,000 studies per year across three hospitals, two outpatient imaging centers, and their own office generates claims where every single study requires a professional-technical component determination. Bill -26 when the practice owns the equipment? That's leaving the technical component on the table. Bill global when reading for an outside facility? That's billing for a technical component the practice doesn't own — and that's a compliance violation. Multiply that modifier decision by 150,000 studies and the margin for error is zero.

The complexity compounds in interventional radiology, where a single procedure can generate 5–10 separate line items — catheter placement codes through selective hierarchy levels, bundled versus separately billable imaging guidance, diagnostic angiography that may or may not be separately reportable depending on whether it's in the same vascular territory as the intervention, contrast administration, and supervision-and-interpretation charges. Then add cross-sectional imaging code selection (without contrast versus with contrast versus with-and-without — a distinction worth $80–$150 per study), mammography screening-versus-diagnostic classification, and imaging guidance charges that most practices leave unbilled because the documentation requirements are a separate step nobody remembers to complete.

AI agents built for radiology billing make the correct modifier, code, and component decision on every study — from the moment the PACS protocol fires to the claim hitting the clearinghouse — recovering the $500K–$1.5M that mid-to-large radiology practices lose annually to systematic billing errors they don't even know they're making.

Professional-Technical Component Billing: The Modifier That Makes or Breaks Radiology Revenue

Every imaging study in radiology has two billable components: the professional component (the radiologist's interpretation and report, modifier -26) and the technical component (the equipment, technologist, supplies, and facility, modifier -TC). When the same entity owns the equipment and provides the interpretation, the study is billed globally with no modifier. When different entities provide the professional and technical components, each bills their respective modifier.

This sounds straightforward until you realize that a typical radiology group reads studies across multiple sites with different ownership arrangements — sometimes in the same day, sometimes for the same patient:

AI agents cross-reference the service location, equipment ownership, and reading physician for every study to automatically apply the correct modifier. When a radiologist reads a chest CT performed at the group's outpatient center (global billing) in the morning and a chest CT performed at the hospital across the street (professional component only) in the afternoon, AI ensures each claim carries the correct modifier despite both studies having the same CPT code, same reading physician, and same date of service.

The Multi-Site Reading Problem

Large radiology groups that cover multiple hospitals and imaging centers face the highest component billing error rates because their PACS worklists blend studies from different facilities with different billing arrangements. A radiologist sitting at one workstation reads 60–80 studies per shift from three or four different facilities — each with its own modifier requirement. Manual billing teams that assign modifiers based on which facility the study came from get it right most of the time, but "most of the time" across 150,000 annual studies means thousands of modifier errors per year.

AI agents eliminate this by maintaining a real-time facility-modifier map that links every PACS source to its correct component billing arrangement. When the group adds a new hospital contract or when a facility transitions equipment ownership (the imaging center the group used to co-own is now independently owned), AI updates the modifier logic across all downstream claims — a change that would take a manual billing team weeks to implement correctly and months to catch the errors from.

Cross-Sectional Imaging: Contrast Protocol Determines the Code and the Payment

CT and MRI studies follow a three-tier coding structure where the contrast protocol determines the CPT code — and the reimbursement difference between tiers is significant:

The "with and without contrast" code (the highest-reimbursing tier) requires that both pre-contrast and post-contrast series were acquired and interpreted. The reimbursement difference between "without contrast" and "with and without contrast" ranges from $80 to $200 per study depending on the body region and payer. For a group performing 50,000 cross-sectional studies per year, even a 5% rate of incorrectly downgrading "with and without" studies to "without contrast" because the billing team didn't verify the contrast protocol represents $200K–$500K in annual revenue loss.

AI agents read the PACS protocol metadata, contrast administration records, and the radiologist's dictated report to confirm which contrast tier was actually performed and select the corresponding CPT code. When a technologist runs a CT abdomen/pelvis with oral and IV contrast (74178 — with and without), AI ensures the claim doesn't default to 74176 (without contrast) just because the ordering physician's original request said "CT abdomen without" and was subsequently upgraded by the radiologist at the time of the study.

Multi-Body-Region Studies and Code Stacking

Combined studies — CT chest/abdomen/pelvis, MRI brain and orbits, CT head and neck — require correct code stacking where each body region is billed as a separate line item with appropriate modifiers. A CT chest/abdomen/pelvis with IV contrast generates three separate codes: 71260 (CT chest with contrast), 74177 (CT abdomen with contrast), and 72193 (CT pelvis with contrast) — or, more commonly, 71260 plus 74178 (CT abdomen and pelvis with and without contrast) depending on the actual protocol. AI validates that the number of billed body regions matches the number of anatomically distinct regions actually imaged and interpreted, preventing both underbilling (billing one code for a multi-region study) and overbilling (stacking codes for regions not separately imaged).

Interventional Radiology: Where Every Procedure Is a Multi-Component Billing Event

Interventional radiology procedures generate the highest per-encounter revenue in the radiology practice — and the most complex claims. A single hepatic chemoembolization procedure can involve selective catheterization through multiple arterial orders, diagnostic angiography of the celiac and hepatic arteries, embolization with both particles and coils, fluoroscopic and cone-beam CT guidance, and contrast administration — each with its own CPT code, bundling rules, and modifier requirements.

Selective Catheterization Hierarchy

Vascular catheterization follows a strict order-based hierarchy where each successive level of selectivity bills an increasingly reimbursable code:

• 36200 — Aortic catheter placement (non-selective)
• 36245 — Selective first-order arterial catheterization (e.g., celiac artery)
• 36246 — Selective second-order (e.g., hepatic artery from celiac)
• 36247 — Selective third-order (e.g., right hepatic artery from proper hepatic from celiac)
• 36248 — Additional second-order or higher selection in the same vascular family

The coding rule: bill the highest order of selectivity achieved within each vascular family. If the interventionalist catheterizes the celiac artery (first-order), advances into the proper hepatic (second-order), then into the right hepatic (third-order), the claim reports only 36247 for that family — the first and second order are included. But if the interventionalist also catheterizes the left hepatic artery as an additional second-order branch, that generates 36248 as an add-on code. AI agents track every catheter position documented in the procedure report and map it to the correct hierarchical code, capturing add-on codes for additional branches that manual coders frequently miss.

Diagnostic Angiography Bundling Rules

Diagnostic angiography (75710 for unilateral extremity, 75716 for bilateral extremity, 75726 abdominal aorta, 75774 each additional vessel) is bundled into interventional codes when the angiography and intervention occur in the same vascular territory. If a diagnostic angiogram of the right iliac system is performed and then angioplasty of the right iliac artery follows, the diagnostic angiography is bundled — not separately billable.

But when diagnostic angiography is performed in a different vascular territory from the intervention, it is separately billable with modifier -59 or an X-modifier. AI agents compare each diagnostic angiography code against each interventional code on the same claim, determine whether they represent the same or different vascular territories, and either bundle the diagnostic or add the appropriate modifier for separate billing. This territory-matching logic is where most manual coding errors occur in interventional radiology — either leaving separately billable diagnostic angiography unbilled or inappropriately unbundling same-territory studies.

Embolization and Multi-Component Procedures

Embolization procedures use a four-code structure based on the clinical indication:

• 37241 — Arterial embolization, non-CNS, non-hemorrhagic (e.g., tumor embolization, UAE)
• 37242 — Arterial embolization, non-CNS, hemorrhagic or traumatic
• 37243 — Arterial embolization for organ ischemia (e.g., renal or splenic)
• 37244 — Arterial embolization, CNS (intracranial)

AI agents select the correct embolization code based on the documented clinical indication, then layer in the catheterization hierarchy, any separately billable diagnostic angiography, imaging guidance charges, and contrast administration. For a practice performing 30–60 embolization procedures per month, systematic capture of every billable component adds $150K–$300K annually over what manual coding produces.

Mammography: Screening vs Diagnostic Classification and Tomosynthesis Capture

Mammography billing errors cost radiology practices more than any other single modality because of the volume — 15,000–30,000 mammograms per year for a typical breast imaging center — and the financial consequence of misclassifying screening versus diagnostic studies.

Screening mammography (77067 — bilateral with computer-aided detection) applies when the patient has no symptoms and no prior abnormal findings requiring follow-up. The patient owes zero cost-sharing under ACA preventive care mandates. Diagnostic mammography (77065 unilateral, 77066 bilateral) applies when the patient has symptoms (palpable lump, nipple discharge), a prior abnormal screening result, or a personal history of breast cancer. Diagnostic mammograms have standard cost-sharing — deductibles, copays, and coinsurance apply.

Misclassifying a screening mammogram as diagnostic shifts an unexpected bill to the patient and triggers patient complaints. Misclassifying a diagnostic mammogram as screening underreports the clinical indication and may result in payer recoupment. AI agents evaluate the ordering indication, prior mammography results in the patient's imaging history, and the radiologist's assessment to classify each study correctly — and flag cases where the classification changed mid-study (a screening mammogram that became diagnostic when the radiologist recalled the patient for additional views).

Tomosynthesis and Add-On Codes

When 3D tomosynthesis (digital breast tomosynthesis) is performed in addition to standard 2D mammography, the tomosynthesis component is billed as an add-on code: 77061 for unilateral or 77062 for bilateral. These add-on codes are missed at a rate of 8–15% across the industry because they require a separate line item that isn't auto-populated in many billing systems. At an average reimbursement of $40–$65 per add-on across 20,000 annual mammograms, missed tomosynthesis codes cost a practice $64K–$195K per year.

AI agents verify that every mammogram performed on a tomosynthesis-capable unit has the corresponding add-on code included on the claim — and that the add-on code matches the laterality of the primary mammography code.

Imaging Guidance: The Most Frequently Missed Charge in Radiology

Imaging guidance codes — fluoroscopic, CT, ultrasound, and MRI guidance for procedures performed by the radiologist or under radiologic direction — are separately billable when proper documentation exists. The documentation requirements are specific: the guidance modality must be identified, the images must be permanently stored, and a formal interpretation of the guidance images must be documented. When any of these elements is missing, the charge cannot be billed.

AI agents check every procedure claim for an accompanying guidance code, verify that CCI edits do not bundle the guidance into the primary procedure, and confirm that the required documentation elements (stored images, formal interpretation) exist in the report. When guidance is performed but not documented separately — the radiologist dictated the biopsy report but didn't include a separate guidance interpretation statement — AI flags the case for documentation completion before billing. For a practice performing 200–400 image-guided procedures per month, systematic guidance code capture recovers $80K–$200K annually.

Contrast Administration and Supply Coding

When the radiologist or radiology staff personally administers IV contrast for CT or MRI studies — as opposed to the facility administering it as part of the technical component — the contrast injection may be separately billable as a drug administration service. In the outpatient setting where the radiology group owns the contrast supply and administers it, the contrast material itself (HCPCS A-codes or Q-codes per milliliter or per dose) and the administration (96374 IV push, 96360/96361 for IV infusion) are chargeable.

AI agents determine whether contrast administration is separately billable based on the service location, equipment ownership, and payer rules — and when it is, calculate the correct contrast material units and administration code. For practices that own and operate their own outpatient imaging centers, contrast supply charges across 30,000–50,000 annual contrast-enhanced studies represent a revenue stream of $100K–$250K per year that practices frequently leave to the facility side of the billing equation.

The Full Radiology Revenue Cycle — Automated

AI agents don't fix one billing problem — they make the correct decision on every study, every modifier, every component, every guidance code, and every contrast charge simultaneously:

• Professional-technical component splits — Facility-modifier mapping for every study, every site, every reading arrangement — -26, -TC, or global on 100% of claims
• Cross-sectional imaging code selection — Contrast protocol verification against PACS metadata, three-tier code matching, multi-region code stacking validation
• Interventional procedure coding — Catheterization hierarchy capture, vascular territory angiography bundling, embolization code selection, multi-component claim assembly
• Mammography classification — Screening versus diagnostic determination, screening-to-diagnostic conversion tracking, tomosynthesis add-on code capture
• Imaging guidance charges — CCI bundling validation, documentation completeness verification, modality-specific guidance code assignment
• Contrast administration — Supply coding, administration charges, site-of-service billing determination

For radiology practices running 8–20 radiologists across multiple sites, the combined impact is $500K–$1.5M in annual revenue recovery — systematic corrections across hundreds of thousands of studies that individually look like rounding errors but collectively represent the difference between a profitable practice and one that's subsidizing payers.