RadNet connects DeepHealth radiology AI with imaging workflows

RadNet combines an imaging-services business with DeepHealth, its digital-health subsidiary and brand. DeepHealth links clinical AI with image viewing, workflow management and reporting. The useful question for a radiology organisation is whether a particular application fits its clinical purpose and information systems, while preserving the radiologist’s review and responsibility for the final report.

01 / ProductDeepHealth connects the image, worklist and report

RadNet’s Reporting Pro launch identifies DeepHealth as a wholly owned subsidiary and the umbrella brand for its digital-health segment. DeepHealth’s software can be considered by external imaging organisations as well as used within RadNet’s own imaging operations.

The portfolio overview separates enterprise imaging from population-health and clinical-AI solutions. Diagnostic Suite provides the image-management and interpretation workspace; Operations Suite concerns the surrounding operational workflow; Reporting Pro brings findings and report creation together. The individual clinical applications have their own indications and deployment limits.

The Diagnostic Suite page describes a unified worklist, image viewer, structured reporting and AI orchestration. It can be deployed as a standalone environment or alongside existing PACS infrastructure. That is a description of supported integration approaches, not proof that every legacy system will connect without configuration or migration work.

RadNet completed the acquisition of Gleamer in March 2026, adding its radiology AI products to DeepHealth. This expands the portfolio’s scope but does not establish that every acquired application is immediately enabled in every installation. A buyer should still identify the named application, version and local commercial availability.

02 / AudienceRelevant when fragmented imaging tools create a specific bottleneck

A radiology group might want to reduce repeated entry of findings, consolidate worklists or make it easier to inspect AI results within the reporting environment. Those are concrete workflows to evaluate. Choosing the whole portfolio simply because it contains many AI applications risks obscuring which problem the first deployment is meant to solve.

An imaging-centre operator may instead focus on scheduling, registration, resource visibility and hand-offs. The Operations Suite page describes a radiology information system and role-oriented workspaces across those tasks. Operational automation is a different claim from diagnostic accuracy, even when both appear under the same brand.

Aidoc is useful context for clinical AI orchestration and workflow, while GE HealthCare provides an adjacent perspective on imaging equipment and connected clinical software. Compare the named application and installation requirements rather than assuming that a single vendor category covers identical capabilities.

The audience is an organisation with clinical, informatics and operational ownership of the workflow. This is not a consumer diagnostic tool for uploading an image and accepting an automated answer. The relevant product documentation, trained users and local clinical governance are part of any meaningful evaluation.

03 / WorkflowProposed workflow: follow one supported study through reporting

For a proposed pilot, choose one examination type and one clearly specified AI application already authorised for the intended use in the relevant market. Define how the study enters the worklist, who interprets it and where the final report is stored. Use the organisation’s approved evaluation arrangements and appropriately governed data.

Follow the image and its identifiers through the existing PACS or the proposed Diagnostic Suite environment. Check that prior studies, patient identity and study context remain associated correctly. An apparently accurate AI finding is not useful if it is attached to the wrong examination or presented without the information the radiologist needs.

Next, inspect the AI output at the point where the radiologist works. Compare the displayed finding or measurement with the source image and the application’s intended use. Include a case where the application produces no output, is unavailable or returns a result that the reviewer disagrees with. The workflow should make those states distinguishable.

The Reporting Pro page describes draft impressions, structured templates, auto-populated findings and quality checks before sign-off. In the proposed evaluation, trace a measurement into the draft and deliberately revise it. Check that the final report reflects the radiologist’s correction and does not retain conflicting text elsewhere.

Finally, evaluate the checked result rather than raw generation speed. Record preparation time, review effort, correction types and the handling of incomplete studies. These are proposed operational measures; they are not clinical findings from Sequenced. No patient images, reporting installation or diagnostic performance were tested for this article.

04 / PricingSpecify the application, integration and market in the proposal

DeepHealth’s public pages provide a demonstration and commercial-contact route. No universal numerical price for Diagnostic Suite, Reporting Pro or all clinical AI applications was verified. A proposal needs to identify the supplied products, deployment arrangement and integration responsibilities, rather than quoting only the name of the wider portfolio.

The June Reporting Pro announcement states commercial availability in the United States and United Kingdom. It describes Australia, South Africa and selected European markets as planned expansion by the end of 2026. At the consultation date, that announcement alone does not prove those later launches occurred; confirm them locally.

Clinical-AI eligibility is also application-specific. The May regulatory update reports particular CE markings and FDA clearances for named Neuro, Prostate and LumbarMR products. A milestone for one application is not a blanket clearance for every feature in the suite or permission for a different clinical purpose.

When comparing proposals, separate image management, reporting, clinical applications, data migration and ongoing operation. Ask what happens if an application is unavailable or a legacy system changes. Those are practical boundaries of the purchased workflow, and they cannot be resolved from a headline about an integrated platform.

05 / DistinctionsIntegration can make an AI result easier to inspect and use

The interesting proposition is the connection between a clinical AI output and the place where it is reviewed. An isolated result creates another item to find and interpret; a linked finding in the worklist, viewer and report can be assessed in context. The buyer should test that traceability rather than assume that integration automatically improves clinical judgement.

Reporting Pro is especially relevant to this hand-off because its described workflow combines dictation, findings, preliminary impressions and quality checks. That is different from presenting a general chatbot beside a radiologist. The intended benefit is reducing disconnected documentation work while retaining a reviewable clinical report.

RadNet’s imaging-services operation also gives the company an internal setting in which to deploy its software. That can inform product development, but company descriptions of internal use remain vendor evidence. This review does not treat a deployment announcement or testimonial as independent proof that another organisation will achieve the same performance.

06 / QuestionsA portfolio-wide claim is not a product-specific evaluation

The first unresolved question is scope. DeepHealth’s pages group several applications into suites, and product disclaimers note that functionality varies by country. Ask for the version, supported modality, intended clinical use and current local availability of each component in the proposed installation. A recognisable suite name is not enough.

The second question is evidence transfer. An accuracy or time-saving result can depend on study design, patient population, readers and comparator. The regulatory update and acquisition announcement include performance and scale claims, but this article does not repeat them as general guarantees. A clinical team should examine the evidence matched to its actual application and setting.

The third question concerns failure visibility. An AI service can be delayed, absent or unsuitable for a study even while the surrounding reporting system remains operational. A useful deployment should make that distinction clear to the reader. The organisation needs to know whether a report was created with a completed AI analysis or without one.

Data governance and system continuity also require a local assessment. Public cloud and interoperability descriptions do not establish the exact hosting, retention, backup or access arrangements in a contract. This review did not inspect private agreements or security evidence and does not infer those controls from the product’s marketing language.

07 / DecisionStart with one reporting hand-off that clinicians can verify

RadNet is a prominent AI-related company because DeepHealth connects a substantial radiology AI portfolio to clinical and operational software. The most useful first decision is still narrow: select an eligible application, define the reporting hand-off and preserve the ability to inspect and correct the result.

Expand only when the organisation understands the local workflow and the evidence supporting it. A connected platform can simplify how information reaches a clinician, but the final value depends on the right product, an appropriate use and an accountable review process.