NMPA granted Innovation Approvals of two diabetic retinopathy (DR) assistant diagnosis software, to Shenzhen Guiji and Shanghai Yingtong. It is the first time that AI technology is approved for ophthalmology diagnosis in China.
China authorities are accelerating the formulation of AI regulations. The latest are below draft standards published by NIFDC (National Institute of Food and Drug Control) on July 2, 2020. Pplease email info@ChinaMedDevice.com for English copies. We charge nominal fees for translation.
- Artificial Intelligence Medical Device Quality Requirements and Evaluation Part 1: Terminology
- Artificial Intelligence Medical Device Quality Requirements and Evaluation Part 2: General Requirements for Data Sets
What Are the Devices
The approvals are based on the neural network of the independent design network structure, through the classification of the fundus image data, the algorithm model training and verification. By obtaining color images of the fundus and using deep learning algorithm to calculate and analyze the images, the product draws the assistant diagnosis recommendations for DR, which are provided to qualified clinicians for diagnosis reference.
Diabetic Retinopathy in China
According to the Global Diabetes Survey (9th Edition) released by the International Diabetes Federation in November 2019, the total number of people with diabetes in China as of 2019 is approximately 116.4 million. DR is the most common complication of diabetes, and about one-third of diabetics have DR symptoms, and one-third of them may have vision-threatening retinal diseases.
Screening is mainly carried out by fundus picturing and then doctor’s reading, while clinical practices such as patient assessment, follow-up check and prognostic evaluation require significant medical resources. the number of professional ophthalmologists in China is insufficient and unevenly distributed, and vast majority of diabetic patients fail to do retinal testing in time. Fundus diagnostics software enables quicker screening and more efficient diagnosis.
AI General Guideline & COVID Triage Guideline
Since April 2018 when the FDA approved the first AI software to detect diabetic retinopathy, NMPA has been busy in preparing the regulatory establishments for AI devices.
NMPA issued the “Technical Guideline on AI-Aided Software” on June 28, 2019. The guideline lists four focal considerations in NMPA review process:
- Needs analysis
- Data collection
- Algorithm design
- Verification and validation
Under the Covid-19 pandemic, NMPA published the “Review Points of CT Image-Assisted Triage and Evaluation Software for Covid-19 Infected Pneumonia” on March 5, 2020. The document provides guidance to AI software manufacturers on the application scope, basic requirements, risk management, software research, clinical research, and software renewal.
Considering that many manufacturers will go through the Innovation Pathway, NMPA is to publish the Technical Review Guideline on AI-Aided Software for Innovation Approval this year.
According to a consulting firm Marketing to China, the domestic market for AI medical devices reached 20 billion yuan ($2.95 billion) in 2018. Accelerated efforts from NMPA will compel China to be the leader of AI applications.