Deepwise Secures NMPA Class III Approval for Pneumonia AI Diagnostic Software

On April 1, the National Medical Products Administration (NMPA) officially granted Class III medical device approval to DeepWise’s “AI-Assisted Triage and Assessment Software for Pneumonia CT Imaging,” which is based on deep learning technology. This marks one of the first pneumonia AI products in China to receive NMPA Class III certification. Just four months after its previous approval, DeepWise has secured another significant milestone.

Founded in 2017, DeepWise leverages deep learning technology and its self-developed artificial intelligence algorithms to continuously expand innovative AI applications in the healthcare sector, gradually building an AI ecosystem that encompasses AI-assisted medical diagnosis, intelligent imaging cloud platforms, smart research solutions, and intelligent medical equipment. Its core division, the DeepWise Research Institute, boasts a scale unmatched within the industry. Dedicated to exploring cutting-edge medical technologies, the institute is responsible for the algorithmic research and development of core products, yielding numerous outcomes with both clinical value and technological innovation.

As one of the first developers of AI for pneumonia to obtain a Class III medical device certification, what insights can DeepWise’s breakthrough offer to the industry?

Pulmonary nodules can almost be regarded as the starting point for artificial intelligence, but their application scenarios are relatively limited. In contrast, due to the complex types of pneumonia and the need to integrate patient laboratory monitoring data, AI for pneumonia must analyze data from multiple dimensions, which also endows it with greater value.

After conquering the challenge of pulmonary nodule detection, DeepWise embarked on the exploration of AI for pneumonia. To date, its approved software not only identifies lesions but also provides quantitative analysis of focal areas, thereby establishing higher R&D barriers for DeepWise’s pneumonia AI products.

The outbreak of the COVID-19 pandemic in 2020 accelerated the advancement of DeepWise’s AI solutions for pneumonia. In less than two weeks, building upon its existing whole-lung product capabilities, DeepWise enhanced the detection and computer-aided diagnosis of pneumonia-related imaging findings, rapidly developing an AI product designed for auxiliary monitoring of COVID-19. This solution was promptly deployed to support epidemic control efforts in Hubei Province. Since ground-glass opacities in the lungs of some COVID-19 patients exhibit very low density during the ultra-early and resolving phases, visual identification by radiologists can be challenging. The application of AI enables automatic and precise detection of pulmonary infection signs in COVID-19 patients, thereby reducing the risk of missed diagnoses.

During the darkest hours of the pandemic, CT medical imaging played a pivotal role as a supplement to nucleic acid testing. To better combat COVID-19, DeepWise’s AI-powered pneumonia solution has made positive contributions in three key areas.

First, AI enables early screening. In the early stages of COVID-19 infection, many patients do not exhibit significant clinical symptoms, but imaging features are already present, with faint ground-glass opacities commonly observed at disease onset. The value of AI becomes evident when physicians face a large volume of patients requiring urgent screening. DeepWise’s pneumonia AI product assists physicians in rapidly and accurately locating lesions, providing detailed quantitative analysis data to support diagnostic and decision-making processes.

Secondly, DeepWise’s unique five-follow-up system enables continuous monitoring of patients’ conditions. The average treatment duration for COVID-19 patients is approximately 15 days. Physicians can leverage DeepWise AI’s follow-up data to observe disease progression, assess severity, and provide an intuitive, visualized record of the entire course of the disease.

Third, AI provides clinical evidence for precision therapy. Clinicians adjust diagnostic and treatment plans based on quantitative data provided by AI, particularly in regulating the use of high-risk medications, thereby assisting physicians in exercising caution in drug administration and ultimately achieving precision medicine.

Leveraging these advantages, DeepWise’s AI-powered pneumonia solution can be deployed in a broader range of primary-care-oriented fever clinics and designated treatment facilities, enabling physicians to detect outbreaks at the earliest stages across all levels, deliver more precise therapeutic interventions, and significantly strengthen future public health epidemic prevention and control capabilities.

According to public information, DeepWise’s newly approved AI product for pneumonia consists of web-based software and server software, featuring functions such as deep learning-based triage alerts for COVID-19, measurement, annotation, quantitative analysis, and follow-up.

During the COVID-19 pandemic, DeepWise leveraged its extensive product expertise to respond rapidly, deploying its independently developed artificial intelligence (AI) technologies and products to the front lines at the earliest opportunity. The company donated two sets of its AI-powered pneumonia diagnostic systems to Wuhan No. 9 Hospital and Huoshenshan Hospital, respectively, thereby supporting epidemic prevention and control efforts. AI technology played a crucial role in large-scale imaging screening, treatment processes, and intelligent post-recovery follow-up, empowering physicians to better manage epidemic control measures across different stages of the outbreak.

In the “Evaluation of AI-Assisted Diagnostic Products for Pneumonia Imaging” conducted last year under the guidance of the Department of Science and Technology of the Ministry of Industry and Information Technology, DeepWise’s AI product for pneumonia demonstrated outstanding performance and was awarded a Certificate of Honor. Furthermore, it was recognized as a Demonstration Case of Technological Innovation in Services at the 2020 China International Fair for Trade in Services (CIFTIS), and its core algorithm was selected for the JinYuan List’s 2020 Breakthrough Innovations in Big Data Industry Technologies.

Furthermore, the unique five-time follow-up feature of DeepWise’s AI product for pneumonia can also accumulate clinical data to support scientific research on COVID-19.

DeepWise has collaborated with numerous medical institutions in China, yielding a substantial body of scientific research. Notably, a paper co-authored with Zhongda Hospital Affiliated to Southeast University, which investigated the relationship between the composition, quantity, and distribution of pneumonia lesions and adverse clinical outcomes, was published in Theranostics (Impact Factor: 8.063). In partnership with Guizhou Provincial People’s Hospital, DeepWise released a monograph on multicenter research findings regarding COVID-19. Additionally, in collaboration with Beijing You’an Hospital Affiliated to Capital Medical University, DeepWise conducted a study exploring the role of CT combined with AI in analyzing pulmonary fibrosis in COVID-19 patients, examining the correlation between pulmonary fibrosis and early circulating immune or inflammatory markers.

Approval has long been one of the major challenges facing medical AI. Under the continuous attention and emphasis of the National Medical Products Administration (NMPA), the approval process is being continuously accelerated.

Since July 2019, with the successive release of policies such as the “Review Points for Medical Device Software Assisted by Deep Learning for Decision-Making” and the “Appendix to the Good Manufacturing Practice for Medical Devices: Standalone Software,” the approval process for medical AI has become increasingly standardized.

In early March last year, in response to the COVID-19 pandemic and in accordance with the principles of scientific review and the requirements for ensuring product safety, efficacy, and quality controllability, the National Medical Products Administration (NMPA) and the Center for Medical Device Evaluation (CMDE) issued the "Key Points for the Review of AI Software for Triage and Assessment of Pneumonia Based on CT Imaging (Trial)." This document outlined the approval process for AI software during the pandemic across seven aspects: scope of application, basic requirements, risk management, software research, clinical studies, labeling, and considerations for software updates. Consequently, the safety validation of AI products for pneumonia diagnosis has been brought onto the right track.

Medical AI has regained its vitality, moving from the winter of 2019 to the successive approval of products. Since 2020, multiple medical AI products in China have obtained Class III medical device certifications, covering fields such as pulmonology, orthopedics, ophthalmology, and cardiovascular care. As increasing capital flows into leading enterprises, the future medical AI market will continue to undergo significant changes.

Regulatory approval is the stepping stone for the commercialization of medical AI products, but it is not the sole factor constraining their development. The key to long-term success lies in whether the product can adapt to the usage habits of hospitals and physicians, meet clinical needs, and align with hospital workflows.

With technological advancements, the continuous expansion of medical data, and the ongoing improvement of hardware capabilities, the integration of AI into healthcare will become increasingly diverse in both its methods and applications. In the future, companies like DeepWise will continue to expand their product offerings with a spirit of innovation, striving to drive industry development.