https://dexpra.com Healthcare interoperability, data platforms, and applied AI. en-CA Wed, 21 Oct 2026 12:00:00 GMT https://dexpra.com/insights/clinical-ai-monitoring https://dexpra.com/insights/clinical-ai-monitoring Wed, 21 Oct 2026 12:00:00 GMT A clinical model is validated once. After go-live its patients, devices and users change, and the organisation running it is usually the only one placed to notice. What to watch, and what it costs. Applied AI Data platforms https://dexpra.com/resources/clinical-ai-monitoring https://dexpra.com/resources/clinical-ai-monitoring Wed, 21 Oct 2026 12:00:00 GMT What changes after a clinical model goes live, who is asked to watch it, the signals and what each costs, how long a fall takes to see, and what the platform has to record. Applied AI https://dexpra.com/resources/healthcare-data-platform-architecture https://dexpra.com/resources/healthcare-data-platform-architecture Thu, 15 Oct 2026 16:38:07 GMT What a healthcare data platform is made of: six components, two shared services, the standards that govern each, and the four questions that decide its shape. Data platforms https://dexpra.com/insights/healthcare-data-platform-architecture https://dexpra.com/insights/healthcare-data-platform-architecture Thu, 15 Oct 2026 12:00:00 GMT The same two words describe an integration engine and a cloud lakehouse. Naming the parts, six components and two shared services, is what makes two very different platforms comparable. Data platforms https://dexpra.com/resources/on-prem-hybrid-clinical-ai https://dexpra.com/resources/on-prem-hybrid-clinical-ai Sat, 10 Oct 2026 18:32:00 GMT What crosses a hospital’s site boundary when the model runs on local hardware: five topologies, eleven paths, the standards on each, and the rules that follow. Data platforms https://dexpra.com/insights/on-premises-hybrid-clinical-ai https://dexpra.com/insights/on-premises-hybrid-clinical-ai Sat, 10 Oct 2026 12:00:00 GMT Running the model on hospital hardware answers latency, bandwidth and residency. It does not stop the data moving. The inference stays; the evidence, the updates and the accountability travel. Data platforms https://dexpra.com/resources/ai-agents-on-fhir https://dexpra.com/resources/ai-agents-on-fhir Sun, 04 Oct 2026 02:28:56 GMT How an autonomous agent reaches a patient record: the three SMART subjects, the two-token bridge, and the audit question nobody can answer afterwards. Applied AI https://dexpra.com/resources/us-medical-ai-regulation https://dexpra.com/resources/us-medical-ai-regulation Sat, 03 Oct 2026 22:17:55 GMT A 33-page map of US regulation for AI in medical software and hardware, and the four rules that are not the FDA’s. Applied AI https://dexpra.com/insights/ai-agent-is-a-client-not-a-user https://dexpra.com/insights/ai-agent-is-a-client-not-a-user Sat, 03 Oct 2026 12:00:00 GMT SMART defines three subjects and an agent is none of them. What a record actually sees when a model asks it a question, and why the answer is usually one service account doing everything. Applied AI https://dexpra.com/insights/clinical-ai-regulation-intended-use https://dexpra.com/insights/clinical-ai-regulation-intended-use Fri, 02 Oct 2026 12:00:00 GMT Intended use decides, not the architecture and not the accuracy. FDA rewrote the four criteria in January 2026, and three of them now read differently. Applied AI https://dexpra.com/resources/imaging-ai-standards https://dexpra.com/resources/imaging-ai-standards Fri, 18 Sep 2026 04:37:39 GMT A 26-page audit of every standard governing AI in medical imaging, and what each one declines to cover. Applied AI https://dexpra.com/insights/imaging-ai-six-handoffs https://dexpra.com/insights/imaging-ai-six-handoffs Thu, 17 Sep 2026 12:00:00 GMT An imaging AI deployment is not one integration. It is six, each governed by a different standard, and the two nobody scopes are the last two: what the radiologist did with it, and who pays. Applied AI https://dexpra.com/resources/mado-101 https://dexpra.com/resources/mado-101 Sat, 12 Sep 2026 22:07:43 GMT A 21-page guide to IHE Radiology MADO for engineers and architects, checked against the supplement Rev. 1.2, 20 August 2026. Interoperability https://dexpra.com/insights/mado-imaging-manifest https://dexpra.com/insights/mado-imaging-manifest Sat, 12 Sep 2026 12:00:00 GMT IHE Radiology published MADO in March and revised it in August. It takes the retrieve address out of the signed report, gives the manifest two encodings, and asks less of a server than XDS-I.b did. IHE https://dexpra.com/insights/data-platform-evidence-package https://dexpra.com/insights/data-platform-evidence-package Wed, 09 Sep 2026 12:00:00 GMT Of 691 AI devices the FDA cleared through July 2023, 95.5% never said who was in the training data. That is not a paperwork failure. It is a platform that could not produce the answer. Data platforms https://dexpra.com/resources/dicomweb-101 https://dexpra.com/resources/dicomweb-101 Sun, 06 Sep 2026 23:13:23 GMT A 23-page orientation to DICOMweb for engineers and architects, checked against DICOM PS3.18, edition 2026c. Interoperability https://dexpra.com/insights/dicomweb-three-services-four-gaps https://dexpra.com/insights/dicomweb-three-services-four-gaps Sun, 06 Sep 2026 12:00:00 GMT Three services, ordinary HTTP, JSON on the wire. DICOMweb is the most approachable thing in medical imaging, and the four things it declines to specify are where the schedule goes. Interoperability DICOM https://dexpra.com/resources/fhir-101 https://dexpra.com/resources/fhir-101 Fri, 04 Sep 2026 18:58:44 GMT A 26-page orientation to FHIR for engineers, checked against R4, R5, SMART App Launch 2.0.0 and US Core STU 9. Interoperability https://dexpra.com/insights/fhir-the-api-is-the-easy-part https://dexpra.com/insights/fhir-the-api-is-the-easy-part Fri, 04 Sep 2026 18:00:00 GMT A FHIR server answers its first request in an afternoon. Profiles, terminology and the implementation guide are the actual project, and none of them is the API. Interoperability FHIR HL7 https://dexpra.com/resources/hl7-101 https://dexpra.com/resources/hl7-101 Fri, 04 Sep 2026 12:45:59 GMT A 24-page orientation to HL7 v2, v3, CDA and FHIR, checked against v2.9.1, C-CDA Edition 5 and FHIR R4. Interoperability https://dexpra.com/insights/hl7-four-standards-one-name https://dexpra.com/insights/hl7-four-standards-one-name Fri, 04 Sep 2026 12:00:00 GMT Ask three engineers in a hospital what HL7 is and you get three answers, all correct. Pipe-delimited text, an XML document, a JSON API: four generations, one name, none switched off. Interoperability HL7 https://dexpra.com/resources/dicom-101 https://dexpra.com/resources/dicom-101 Mon, 31 Aug 2026 22:24:25 GMT A 32-page orientation to DICOM for engineers and data scientists, checked against the current edition, PS3 2026c. Interoperability https://dexpra.com/resources/ihe-101 https://dexpra.com/resources/ihe-101 Mon, 31 Aug 2026 22:24:25 GMT A 29-page walk through IHE Radiology and Scheduled Workflow.b, transaction by transaction. Checked against RAD TF Rev 23.0. Interoperability https://dexpra.com/insights/two-conformant-systems-one-broken-department https://dexpra.com/insights/two-conformant-systems-one-broken-department Mon, 31 Aug 2026 13:00:00 GMT DICOM and HL7 define the words. They do not define the conversation. IHE writes the conversation down, and Scheduled Workflow.b is the one running almost every radiology department. Interoperability IHE https://dexpra.com/insights/dicom-is-not-a-file-format https://dexpra.com/insights/dicom-is-not-a-file-format Mon, 31 Aug 2026 12:00:00 GMT Every scanner, archive and viewer on earth speaks DICOM. Almost every team building AI on medical images discovers, painfully, that it is a four-layer stack and not a header. Interoperability DICOM