Register for Upcoming Webinar on DEC. 8 @ 11AM

2024 China NMPA Bluebook is here:

AI Infused Software as a Medical Device (SaMD) for Ischemic Stroke Detection Approved in China


The NMPA granted innovation approval to Ischemic stroke CT image assisted assessment software and released the review report in July 2023.

This software is designed to aid healthcare professionals in evaluating and categorizing abnormal CT scans of the brain. Employing AI algorithms to analyze CT image data and generate Alberta Stroke Program Early CT Score (ASPECTS), it quantitatively assesses the suspected acute large artery occlusion ischemic stroke patients within onset of 6 hours.

Product Overview, Pre-clinical and Clinical

As NMPA standardizes and streamlines the review process for fast-track approval, it is vital for manufacturers to comply with following submission materials requirements if you have similar software entering China market.

Product Overview

  1. Product structure and composition
  2. Scope of product application
  3. Model/Specification
  4. Working principle

This product uses deep learning algorithm to realize automatic brain region segmentation based on head CT plain scan image data and determines whether the score of the scoring area is deducted, so as to obtain the ASPECTS score of the case.


  1. Product performance research
  2. Packaging and shelf-life research
  3. Software research

The software security level is medium. The applicant submitted the software research data according to the “Guideline for Medical Device Software Registration Review (2022 revision)”, which confirmed that the design and development process of the product software are controllable, and the comprehensive residual risks are acceptable.

The cybersecurity research data was submitted according to the “Guideline for Cybersecurity Review of Medical Devices (2022 revision)”, confirming that the existing cybersecurity risks of the product are controllable, and formulating an emergency response plan for cybersecurity events.

The research data of deep learning algorithm was submitted according to the “Guideline of Artificial Intelligence Medical Device”, including data diversity analysis, data labeling quality control, training data-evaluation index curve, algorithm performance evaluation and other research data, which confirmed that the performance of the software algorithm could meet the design requirements.

Clinical Evaluation

The applicant conducted retrospective, multi-reader, multi-case (MRMC) clinical trials.

The primary outcomes were as follows: the consistency of early ischemic stroke ASPECTS score and expert score was compared between the experimental group and the control group.

The secondary outcomes were the correlation of ASPECTS scores between the experimental group and the control group with the expert group (case as unit), and the consistency of each brain region (different brain regions as unit). The consistency and correlation of ASPECTS scores independently read by software and expert group were evaluated. The software performance was evaluated from the aspects of functional use, operation convenience and stability, and was divided into three levels: satisfactory, general and unsatisfactory.

The safety index was the incidence of device defects.

The primary outcomes showed that for FAS analysis set and PPS analysis set, the consistency difference between the ASPECTS score of early ischemic stroke a with or without AI and the expert group score (FAS: 8.14%; PPS: 8.18%) lower limit of 95% confidence interval (FAS: 7.51%, PPS: 7.54%) were greater than 0; The lower limit of 95% confidence interval (FAS and PPS: 0.79) between physicians using AI and the expert panel was greater than 0.75, showing good consistency.

The lower limit of 95% confidence interval (FAS and PPS: 0.55) between physicians not using AI and the expert panel. 0.53) was greater than 0.4, the results were basically consistent. The consistency between the radiologist using AI and the expert panel was higher than that without AI. The results of each radiologist had the same conclusion.

For detailed information on the clinical trial data, please email us at We can translate for you with nominal fees.

How Common Strokes Are in China

Since 2015, stroke has become the leading cause of death and disability in China, posing a significant threat to the health of its citizens as a major chronic non-communicable disease.

According to the China Stroke High-risk Population Screening and Intervention Program, an estimated 17.8 million adults in China had experienced a stroke in 2020, with 3.4 million experiencing their first-ever stroke and another 2.3 million dying as a result. Additionally, approximately 12.5% of stroke survivors were left disabled, as defined by a modified Rankin Scale score greater than 1, equating to 2.2 million stroke-related disabilities in 2020.

As the population ages and the prevalence of risk factors like diabetes, hypertension, and hyperlipidemia continues to rise and remains poorly controlled, the burden of stroke in China is increasing.

Related Posts