Skip to main content

Software and SaMD Classification

When software qualifies as a medical device​

Software (including mobile applications and AI/ML systems) is regulated as a Programme Medical Device (プログラム医療機器) in Japan if it meets the Article 2 definition — that is, if it is intended for use in the diagnosis, treatment, or prevention of disease, or to affect the structure or function of the human body, and is listed in the Cabinet Order.

The 2014 PMD Act amendments and subsequent MHLW notifications (most notably the 2019 notification on Programme Medical Devices) created a dedicated framework for SaMD. The key MHLW notification distinguishes between:

  1. Software that is itself a medical device — regulated as a Programme Medical Device
  2. Software that controls or operates a regulated medical device — regulated as part of or an accessory to that device
  3. Software that merely supports a regulated device (data storage, communication) — typically not itself a medical device

Classification of Programme Medical Devices​

Once software is determined to be a Programme Medical Device, it is classified using the same JMDN-based system. MHLW has created specific JMDN codes for software categories. The class (I–IV) depends on the intended diagnostic or therapeutic function:

  • Class I: General wellness software or low-risk tools with no significant clinical decision-making
  • Class II: Diagnostic aids that support (but do not replace) clinical judgment for non-serious conditions
  • Class III: Diagnostic or treatment support software for serious conditions; software driving active therapy devices
  • Class IV: Software making autonomous critical clinical decisions; software for blood safety applications

AI/ML medical devices​

MHLW issued guidance in 2021 on the regulation of AI-based medical devices. Key points:

  • AI/ML software may be a Programme Medical Device if it meets the intended use criteria
  • The classification follows the same JMDN-based approach
  • Change management is a key concern: AI/ML systems that update their algorithm post-approval may trigger Ichibu Henko (partial change approval) obligations depending on how the learning affects performance

Cybersecurity​

MHLW and PMDA have published guidance on cybersecurity requirements for networked and software-enabled devices. Cybersecurity documentation is expected as part of the pre-market dossier for connected devices and Programme Medical Devices.