PMDA Updates Ultrasound Metrics Validation Guidelines with AI Measurement Limits

On May 7, 2026, the Pharmaceuticals and Medical Devices Agency (PMDA) of Japan revised its Ultrasound Metrics Device Performance Verification Guidelines, introducing mandatory measurement accuracy thresholds for portable ultrasound devices incorporating AI-assisted measurement functions. This update directly affects manufacturers exporting to Japan—particularly those in the medical imaging equipment sector—and signals tightening regulatory expectations for algorithmic performance validation in clinical diagnostic tools.

Event Overview

On May 7, 2026, the Pharmaceuticals and Medical Devices Agency (PMDA) published an updated version of its Ultrasound Metrics Device Performance Verification Guidelines. The revision introduces a new mandatory requirement—the ‘measurement bias limit’—applicable specifically to portable ultrasound devices equipped with AI-assisted measurement capabilities. Under the updated guidelines, devices must meet the following performance criteria: (1) distance measurement error in B-mode ≤ ±0.5 mm; (2) blood flow velocity measurement error ≤ ±8%. Additionally, manufacturers must submit a third-party algorithm validation report certified to JIS Z 8401. As confirmed in public reports, three leading Chinese ultrasound exporters have suspended shipments to Japan pending compliance assessment.

Industries Affected by the Revision

Direct Exporters of Ultrasound Devices

These companies are directly subject to PMDA’s verification requirements. The new bias limits apply at the device registration and post-market surveillance stages. Impact manifests as delayed market access, halted shipments (as observed with three Chinese firms), and increased pre-submission testing burden—especially for AI modules not previously validated to JIS Z 8401.

Medical Device Contract Manufacturers & OEM Suppliers

Suppliers providing AI software integration, embedded measurement modules, or firmware development services face upstream demand shifts. Clients may request revalidation of existing AI algorithms against the new thresholds, triggering revisions to design history files, traceability documentation, and verification protocols—particularly where JIS Z 8401–aligned test methodologies were not originally specified.

Regulatory Affairs & Clinical Validation Service Providers

Firms offering regulatory strategy, technical file preparation, or third-party algorithm validation services are seeing increased inquiry volume related to JIS Z 8401 alignment. The requirement for JIS Z 8401–certified reports—not just internal or ISO/IEC 17025–accredited testing—introduces a narrow qualification criterion that limits eligible service providers in non-Japanese markets.

Key Considerations and Recommended Actions for Stakeholders

Monitor Official PMDA Communications for Implementation Timelines and Transitional Provisions

The revised guidelines took effect on May 7, 2026—but PMDA has not publicly clarified whether grandfathering applies to already-registered devices, or whether pending applications must comply immediately. Exporters should track PMDA’s official notices (e.g., via the PMDA website’s ‘Guideline Updates’ section) for any grace periods, phased enforcement dates, or Q&A documents addressing legacy AI functionality.

Review AI Algorithm Documentation Against JIS Z 8401 Requirements—Not Just General Accuracy Claims

JIS Z 8401 specifies procedures for evaluating numerical accuracy in measurement standards—including uncertainty estimation, repeatability testing under defined conditions, and reporting formats. Many existing AI validation reports reference ISO/IEC 13485 or IEC 62304 but omit JIS Z 8401–specific statistical rigor. Companies should audit current validation packages for alignment with JIS Z 8401’s clauses on bias quantification, confidence interval reporting, and test phantom specifications.

Assess Supply Chain Dependencies for Third-Party Validation Capacity

Only laboratories accredited to JIS Z 8401 (not merely ISO/IEC 17025) can issue compliant reports. Such labs are currently concentrated in Japan and a limited number in South Korea and Singapore. Exporters relying on non-Japanese labs should verify accreditation scope and lead times—some facilities report >12-week queues for JIS Z 8401–aligned ultrasound algorithm assessments.

Prepare for Potential Re-evaluation of Existing Registered Devices

Although the guideline revision targets new submissions, PMDA retains authority to request post-market performance data under Article 71–2 of the Act on Securing Quality, Efficacy and Safety of Products Including Pharmaceuticals and Medical Devices. Companies with AI-enabled ultrasound devices already on the Japanese market should proactively compile historical measurement accuracy data—especially from clinical use settings—to support potential future inquiries.

Editorial Perspective / Industry Observation

Observably, this revision reflects PMDA’s increasing emphasis on verifiable, standardized performance metrics for AI-enabled medical functions—not just safety and software lifecycle management. It is less a standalone policy shift and more a targeted calibration of existing regulatory expectations toward quantitative, testable outcomes. Analysis shows that the ±0.5 mm and ±8% thresholds align closely with clinical decision-making tolerances reported in Japanese echocardiography and obstetric ultrasound practice guidelines—suggesting the limits are clinically grounded rather than arbitrarily stringent. From an industry perspective, this update is best understood not as an isolated compliance hurdle, but as an early indicator of how major regulators may operationalize the ‘AI-specific performance verification’ principle now emerging across global frameworks (e.g., FDA’s AI/ML Software as a Medical Device framework, EU MDR Annex III updates). Current focus should remain on implementation readiness—not theoretical alignment.

The PMDA’s May 2026 revision marks a concrete step toward enforceable performance accountability for AI-assisted ultrasound measurements. Its significance lies not in novelty alone, but in its binding, numeric specificity and its linkage to a nationally recognized metrological standard. For stakeholders, the most rational interpretation is that this is both a signal and an immediate operational requirement: a signal of escalating global expectations for AI validation rigor, and a requirement demanding near-term attention to documentation, testing infrastructure, and supply chain coordination. Continued monitoring—not delay—is the appropriate response.

Source: Pharmaceuticals and Medical Devices Agency (PMDA), Japan – Ultrasound Metrics Device Performance Verification Guidelines (Revised Edition, Effective May 7, 2026). Publicly confirmed shipment suspensions reported by industry trade publications (as of May 2026).
Note: Ongoing observation is recommended for PMDA-issued transitional guidance, clarification on applicability to legacy devices, and updates to JIS Z 8401 implementation interpretations.