When Ultrasound Records Change: A 2026 GUDID, EUDAMED and 510(k) Support-Planning Audit

What a registry can tell a service team

A distributor may be asked to support a console whose original labeler no longer appears to sell it. A hospital may be considering replacement because a supplier describes the model as “withdrawn.” Before treating either statement as a parts-purchasing signal, separate four questions: what is registered, what is currently sold, what remains installed, and what can be supported under the relevant service arrangements.

GUDID describes device identifiers and associated labeler information. Its commercial distribution status is useful for checking a particular US record. It does not contain a hospital's installed-base census or an OEM's complete end-of-service commitments. A change in record status should therefore open a verification task rather than close a clinical or capital decision. ²⁴

EUDAMED adds another jurisdiction's registration records. A 510(k) record identifies a premarket decision and the submitted device context. Neither is a unit-sales series. Differences among these sources can be informative, but only after the model, legal entity, device identifier and market are matched. ³⁵

The practical output is a model support file: identifiers and name aliases, relevant regulatory records, exact equipment configuration, current OEM documentation, local installed equipment, recent failures, validated repair options and replacement alternatives. Public databases populate only part of that file.

The US selection: records overlap across product codes

We used a broad 15-code selection from the stored GUDID analysis: ITX, IYO, IYN, JAF, HGL, JOP, OQQ, PAA, DXK, ODG, PSV, OBJ, SGJ, SGH and HEM. The narrower comparison uses ITX, IYO and IYN. FDA's classification database identifies the scope of each product code; the broad selection is an operational search definition, not a claim that all its records are diagnostic imaging consoles or reusable probes. ⁶

The individual rows cannot be summed. A record may carry both system and transducer-related codes, so its status appears in more than one product-code query. The union query counts each returned public device record once. The 8.3% core proportion is a cross-sectional status share; it is not an annual attrition rate, lifetime withdrawal probability or estimate that one in twelve installed scanners needs replacement. ¹

A Device Identifier is not a machine serial number

A GUDID record is anchored by its primary Device Identifier and can contain additional identifier information, including higher packaging levels. It is not a row for each installed console. Multiple records may refer to configurations within one commercial family, and names may vary with the labeler or market. Counting unique model-description strings still would not fully normalize those families. ²⁴

Accessories further complicate an ITX-only count. A needle guide, cover or biopsy-related item can appear within a code selection a service engineer associates with transducers. The record description and intended use need to be read before it enters a probe-support list. FDA's ultrasound submission guidance likewise provides device and submission context rather than a one-code-one-probe counting rule. ⁶⁷

A single-use flag can help prioritize review, but it should not be used as proof that every remaining item is an active reusable imaging probe. Conversely, excluding all single-use records without inspecting their intended use can discard relevant products. A defensible support inventory keeps the inclusion and exclusion reason for each matched device family.

Distribution end dates cluster, but they do not explain themselves

The end-date series below uses the deduplicated core union. Across the 883 currently withdrawn ITX/IYO/IYN records, the 2021 count is 319, or 36.1%. This is a distribution of recorded end dates among records that are withdrawn in the queried snapshot. It does not show every historical status transition, and it does not reconstruct the date on which a service provider first learned about a change. ¹

The 2021 concentration deserves investigation. It does not support attributing every withdrawal to a corporate acquisition or assuming that a regulatory update precedes every customer notice. A status record has no field proving the business reason, the last spare-parts shipment or the remaining service commitment.

Hitachi: a concentrated administrative event, with a date mismatch to avoid

The exact labeler “HITACHI, LTD.” has 242 withdrawn ITX records in this query, representing 35.0% of the 692 withdrawn ITX records. Of those 242, 239 show June 30, 2021; three show dates in 2018. Fujifilm's acquisition of Hitachi's diagnostic imaging business completed on March 31, 2021. The acquisition and the large end-date cluster occurred in the same year, but they are not the same date or direct evidence of a common device-support outcome. ¹⁸⁹

The selected labeler bars total 586 of the 692 withdrawn ITX records; omitted labelers account for the remainder. Labeler names are retained as record identities, rather than merged into present-day corporate brands. This prevents an administrative portfolio from being mistaken for a complete brand-wide history. ¹

A current Fujifilm page for ARIETTA Precision provides a useful countercheck to sweeping claims that an acquired ultrasound portfolio has simply ceased to exist. It does not establish support for every older Hitachi configuration; it demonstrates why product continuity has to be checked at the current product and service level. ¹⁰

Carestream: do not attach a later corporate story to an earlier date

The 21 withdrawn Carestream ITX records in this selection have 2020 end dates: five on May 18, eleven on June 8, one on September 24 and four on October 9. A corporate restructuring completed years later cannot be used to explain that cluster without additional evidence. We therefore use the dates as a record-level finding and make no causal claim about the company's subsequent organization. ¹

For either example, the next useful action is to contact the current support organization with the precise identifier, model and configuration. Ask for the applicable support status, documentation revision, parts availability, service conditions and any announced dates. A brand-level headline cannot answer those questions for a particular installed machine.

Publication year measures registration activity, not model age

The stored July 7, 2026 GUDID export contains the same 11,441-record broad selection. Its publication-year distribution includes three records in 2014 and three in 2015, as well as the larger 2016–2026 groups. Including those early records reconciles the chart to the selection total. ¹⁴

The 2016 concentration is consistent with a major Class II UDI compliance milestone, but a publication date does not establish when a model was designed, cleared, first sold or installed. FDA's UDI compliance schedule provides the policy context; it does not justify dating every 2016 record's underlying equipment to 2016. ¹¹

For a replacement assessment, gather the individual machine's manufacture date, installation history, software version and maintenance record. The 510(k) decision and GUDID publication date can be supplementary context. Neither is a substitute for a serial-specific service history or performance assessment.

EUDAMED: the broad keyword result is a search-quality warning

We parsed 2,926,029 JSON records from the July 25, 2026 EUDAMED snapshot. The exploratory filter searches the concatenated trade and device names for ultrasound, ultrasonic, transducer, sonograph, echograph or probe. It returns 8,943 matches. That filter is deliberately broad, and manual checks show that it includes laboratory probe reagents and cleaners, ultrasonic tissue-processing products, procedural instruments and other non-imaging devices. The matches include 228 IVD records. ³¹²

Of these search matches, 8,709 are on the market, 106 are no longer on the market and 128 are not intended for the EU market. The on-market proportion is 97.4%. This result cannot be compared directly with the US core-code withdrawal share: the selection definitions, jurisdictions, status vocabularies and snapshot dates differ. The 1.2% no-longer-on-market share does not prove that European imaging equipment has a lower withdrawal rate. ³

EUDAMED's registration timetable is another source of context. Mandatory use of the UDI/device registration module began on May 28, 2026, with transition provisions for existing devices. That timing may affect the composition of registered records, but it does not independently explain the observed status distribution or supply a missing historical installed-base denominator. ¹²

The manufacturer SRN-country prefix is China for 4,506 matches, or 50.4%, and the United States for 1,790. China is 2.52 times the US count in this search result. The chart includes an Other group of 548 records so that all categories reconcile to 8,943. SRN country is the manufacturer actor's country field; it is not a count of factories, shipments or hospital installations. ³

A service business should not use this broad search to predict that Chinese brands will occupy half its future repair queue. To make an imaging-specific dataset, review the nomenclature and intended purpose, exclude laboratory and surgical matches outside scope, normalize commercial families, and document the retained identifiers. Until that work is done, the broad result is useful for understanding the limitations of a search, not for purchasing spare parts by country share.

510(k): premarket context, with a partial 2026 endpoint

The stored FDA 510(k) extract contains 2,513 distinct K numbers whose primary product code is ITX, IYO, IYN or JAF. It was imported on August 24, 2026, but its latest decision date is June 29, 2026. Calling the series complete through August would confuse the import date with the source's decision coverage. Secondary product-code matches are outside this selection. ⁵

The 2011–2025 series averages 68.9 selected decisions a year. The 34 decisions shown for 2026 cover only the available period through June 29 and should not be annualized as a forecast. These counts do not directly measure new console families, probe innovation, units sold or systems entering service contracts. A single submission can cover related components or configurations, and an ITX primary code can include a biopsy kit. ⁵⁷

One extract row has an uninformative decision label. Its K number, K041638, resolves in the official FDA record to a September 2, 2004 substantially equivalent kit decision. We retain it as a checked decision, rather than inventing an additional outcome category or silently dropping it. ¹³

Applicant country is a historical address field. The US accounts for 1,853 of the 2,513 records, China for 252 and South Korea for 121. This does not establish manufacturing origin, current headquarters, market share or the country producing the fastest-growing installed fleet. Corporate naming and filing structures change over a series extending back to 1977. ⁵

For a model support file, use the K number to locate the cleared device description and available summary, then check its relationship to the exact configuration in front of you. A clearance date is one piece of the equipment's regulatory history, not its remaining useful life.

Turn a registry change into a support work order

The following matrix is a proposed operational framework. It is not a stock-return model fitted to the registry data. Its inputs must come from the service organization, its customers and verified suppliers.

A useful quarterly review starts with the installed equipment under contract, not every row in a regulatory database. Record each system's model and serial, board part numbers and revisions, supported software, probe interfaces and current service responsibilities. Reconcile model aliases before linking those systems to public identifiers.

Then compare the prior and current saved registry responses. Separate a changed status from a changed labeler name, a newly published record or a corrected description. Retain the response date and API update metadata. Where a query returns multiple pages, retrieve every page and deduplicate by the public device record key; a first page of 100 results is not a census. ¹¹⁴

For each material change, request a written support check. Obtain available parts and service terms, identify any end-of-sale or end-of-service notice, and clarify whether the notice applies to the exact configuration and market. Add local failure history, actual turnaround times and replacement options before approving stock expenditure. The quarterly interval is a suggested management cadence, not an FDA requirement or a guarantee of adequate warning.

A candidate part still needs compatibility and acceptance evidence

Before a board or probe enters a support plan, establish its identity and compatible hardware/software revisions. Document its provenance, test conditions, result and acceptance criteria. A catalogue listing or successful repair on another machine is not evidence that a different configuration is interchangeable. Rongtao's parts provenance guide and used-system buying guide provide related procurement questions. ¹⁵¹⁶

The same principle applies to donor systems. Their economics depend on acquisition, transport, testing, storage, usable yield and demand for the recovered parts. This audit supplies no evidence for a universal five-to-ten-year opportunity window, a minimum of two stocked boards, or a fixed point at which harvesting becomes cheaper than repair. Those decisions need a model-specific cost and acceptance assessment.

Commercial withdrawal and servicing rules are separate questions

A distribution status entry is not itself a recall, an end-of-service notice or an assessment of an installed device's safety. It also cannot provide a blanket legal permission to operate or modify equipment. Check relevant recalls, OEM notices, local requirements and the condition of the specific device. ²⁴

FDA's 2024 remanufacturing guidance evaluates activities by their effect on the finished device's performance, safety specifications and intended use. The distinction between servicing and remanufacturing is activity-specific. A commercial withdrawal date does not answer that analysis, and not every component replacement or design-related change has an identical regulatory outcome. Service providers should document their actual activities and assess the applicable obligations. ¹⁷

A support-file template for a real customer enquiry

A useful record should let another engineer reproduce the identification and another buyer understand the commercial decision. The following template is a proposed working document. It deliberately leaves the customer's facts blank; the public registry cannot populate them automatically.

Treat an uncertain identifier match as an unresolved task. Similar model names can hide different revisions, market configurations or accessories. A title-only search result is a lead for investigation; attach the underlying record and explain the match before using its status in the support file. If the relationship remains ambiguous, ask the current manufacturer or responsible support organization for clarification instead of assigning a withdrawal date by analogy.

Keep an original supplier statement with its date and scope. A sales representative's statement that a model is “obsolete” may refer to sales, a software release, a particular part or a service contract. Ask which of those meanings applies and retain the answer. The cost comparison will differ if the constraint is a long replenishment time, an unsupported software configuration or a part that cannot meet the required acceptance criteria.

The decision owner should also distinguish usable inventory from listed inventory. A board awaiting identification, a part reserved for another customer and an untested harvested assembly do not provide the same response capability as an identified, accepted part available for the proposed job. Record those states explicitly when deciding whether a new purchase reduces a service risk or merely adds another unqualified item.

Close the enquiry with a documented outcome: proceed under stated conditions, obtain missing evidence, offer a different support option, or decline because the required configuration or acceptance path cannot be established. Retain the reason and a review date. That record makes the next registry update or repeat customer request easier to assess without turning a database status into a promise the service organization cannot support.

Where Rongtao's catalogue helps, and what it cannot prove

At the September 5 review, Rongtao's public board catalogue contains 254 SKU entries and 143 distinct brand/model-description strings. The latter includes aliases and combined descriptions; it is not 143 independently normalized equipment models. The inventory data contains 316 rows and 276 distinct brand/part-number/model tuples. Those are catalogue records, not a verified physical stocktake or a measure of installed demand. ¹⁸

Hitachi and Aloka account for 18 of the catalogue's 143 description strings. That indicates subjects for a compatibility enquiry; it does not establish current stock of every required revision, brand-wide repairability or the number of hospitals using those systems. A buyer should request availability and test evidence for the exact part and equipment configuration. ¹⁸

The repair cases archive can help identify relevant documented experience. Use a matching case to ask about the fault, configuration, method, acceptance evidence and commercial terms. Do not infer universal turnaround or warranty coverage from a general service description. An enquiry should include the model, part number, revision, software, fault history and required destination so that the response can be specific. ¹⁹

Methodology and boundaries

The source package retains the R2 dataset identifiers, manifest keys, original analysis scripts and rejection records. The stored GUDID export is dated July 7, 2026. Status queries were retrieved on September 5 from openFDA, whose metadata reports August 3 as the update date. Counts are scoped to the documented product-code selections. The independent recheck uses OR queries and verifies unique public record keys for withdrawals; end-date years come from those deduplicated records. ¹¹⁴

For EUDAMED, all 20 compressed parts were checked against their manifest hashes and parsed without JSON errors, yielding 2,926,029 rows. The 8,943-result name filter is reproduced in the analysis script. Its false-positive examples and IVD inclusion are retained because they materially limit interpretation. Country is derived from the manufacturer SRN prefix and status from the snapshot's device-status field. No US-to-EU device-level join is claimed. ³

For 510(k), the unit is a distinct K number selected by the four primary product codes. Decision dates, not import dates, determine annual coverage. Country is the applicant field. The latest available decision is June 29, 2026; K041638 was checked against the official record. The public catalogue counts are a separate company-data snapshot with no claim of physical stock verification. ⁵¹³¹⁸

The analysis does not estimate equipment failure rates, remaining service life, installed fleet size, repair margins or the causal effect of acquisitions. Its useful result is narrower: public registry changes can help prioritize a documented support check, while purchasing and service decisions remain tied to identifiable equipment, current terms and validated capability.