48 AI independent software have been approved so far, according to work summary of Artificial Intelligence Medical Device Innovation Cooperation Platform that NMPA CMDE established in July 2019.
The notice says that data types of approved software cover a variety of fields: imaging, physiological information, and in vitro diagnostics, etc.; Product types cover auxiliary diagnosis, auxiliary testing, auxiliary triage, and assessment. Also, 14 independent software entered the innovation channel, including CT imaging, fundus photography of diabetic retinopathy, digestive endoscopic assisted diagnosis, and surgical planning assistance.
AI Independent Software NMPA Guidelines
8 review guidelines and review points have been released. Among them,
- “Guideline for the Registration and Review of Artificial Intelligence Medical Devices”: specifies the general requirements for AI medical devices in detail.
- “Review Key Points for Deep Learning Assisted Decision-Making Software”: the first in the world to clarify the regulatory focus items of related products, which has aroused widespread international attention.
- “Review Key Points for Pneumonia CT Image-assisted Triage and Evaluation Software (Trial)”: played an important role in the fight against the Covid-19 pandemic.
At present, there are additional 61 standards and technical documents formulated by the platform, some of which were already released.
Databases & Working Groups
The databases built by the platform cover multiple data fields such as fundus photography, pneumonia CT, thyroid ultrasound, etc., which help the development and market approval of AI medical products. There are also databases under construction, such as pathological images of cervical cells and multimodal images of pathological myopia.
- Data Governance Group: data-quality-control-focused study of process, methods and requirements of data governance of AI medical devices.
- Real-World Data Group: use the methodological research to guide the databases, which are used to support product R&D training and performance verification.
- Evaluation technology Group: responsible for quality control and product verification, studies data quality requirements and product evaluation methods.
- Clinical Evaluation Group: research on supplementation of clinical evaluation data to reduce or replace some clinical trial.