Eyesight AI Pioneer Yiwei Technology Files IPO Prospectus: Leveraging Years of Technical Expertise to Overcome Barriers in AI-Driven Ophthalmology

The Yellow Emperor's Inner Canon, the foundational text of traditional Chinese medicine, contains a well-known maxim: "Superior physicians treat disease before it arises; average physicians treat disease after it has arisen; inferior physicians treat disease in its advanced stages." However, due to the scarcity and uneven distribution of medical resources, disease prevention and early screening are often overlooked, with resources disproportionately allocated to treatment. Consequently, the concept of "treating pre-disease conditions" remains an unfulfilled "dream."

The advent of artificial intelligence has organically combined human wisdom with machine efficiency, making the realization of the dream of “preventive healthcare” possible. According to data from VCBeat’s “2018 Medical Artificial Intelligence Report,” lung nodule screening and diabetic retinopathy screening have become the two hottest areas in medical AI. In addition to continuous investments by giants such as Tencent and Google, major startups are also fiercely competing, thereby driving the industry’s development.

After three years of development in 2016, 2017, and 2018, the application of AI in ophthalmology still faces an insurmountable hurdle: how to break through the technological bottlenecks of AI, achieve commercialization of products, and establish unique core competitive advantages in the market and industry. To address this, VCBeat conducted an exclusive interview with Ke Xin, founder and CEO of EVsion, an expert who has been deeply engaged in the field of artificial intelligence in ophthalmology for many years.

Compared to its peers, EVsion is a rising star, yet Ke Xin is a pioneer in the industry and one of the earliest trailblazers in AI-driven ophthalmology. A graduate of the Chinese Academy of Sciences, he joined the R&D Center of Daheng Imaging immediately after graduation to engage in intelligent ophthalmology research and development. Later, as team leader, he assumed full responsibility for product development and promotion. Daheng Imaging was among the first companies in China to establish a presence in the ophthalmology sector, having jointly undertaken multiple ophthalmology-related projects funded by the Ministry of Science and Technology with Beijing Tongren Hospital since 1993.

In 2015, when AI-based ophthalmology was just emerging in China, Ke Xin led his team to collaborate with Director Yang Jinkui of the Endocrinology Department at Beijing Tongren Hospital to jointly develop the “Intelligent Analysis System for Diabetic Retinopathy Screening Based on Computer Vision.” The system was subsequently deployed in dozens of top-tier public and private hospitals across China. In 2018, to further realize the value proposition of “preventive medicine through early intervention,” Ke Xin founded EVsion, aiming to leverage fundus examination as an entry point to usher in an era of AI-driven early disease screening and to promote the development of precision and personalized healthcare.

Led by Ke Xin, the EVsion team has established its foothold through technology and products, with all core members hailing from the Chinese Academy of Sciences and Tsinghua University. EVsion places great emphasis on technology, regarding it as the foundation of an AI company. The company uniquely integrates computer vision and deep learning technologies, offering efficient, low-cost solutions for screening ocular diseases and systemic chronic conditions through its original technical approach. Shortly after its inception, EVsion stood out among more than 40 competitors in the prestigious World Artificial Intelligence Conference (WAIC) technology competition, securing second place in the finals. Currently, EVsion’s products have been piloted in numerous hospitals across China, and the company has established partnerships with industry leaders such as Novartis and Pfizer.

From the perspective of industry demand, the integration of artificial intelligence (AI) into fundus disease screening is not merely an added bonus, but a solution to a critical need. AI will become a powerful tool for alleviating physicians' workload and improving the efficiency of early screening. Furthermore, industry experts consider "AI + Ophthalmology" to be the sector with the greatest potential for large-scale commercialization. However, it is undeniable that significant barriers and challenges remain within AI, particularly in achieving breakthroughs in establishing effective connections between AI systems, physicians, and patients. Compared to other companies in the field, EVsion has taken a distinct approach in this regard.

Leveraging foundational technologies, EVsion comprehensively utilizes computer vision and deep learning techniques to extract lesions from fundus images and quantify feature data. Subsequently, it employs data mining and big data analytics to conduct in-depth analysis of fundus imagery, ultimately developing an AI-assisted diagnostic system for fundus imaging.

Ke Xin also highlighted EVsion’s AI capabilities: “Our AI not only provides conclusions but also annotates key lesions, critical structures, and important locations, while simultaneously acquiring quantitative data—a feature that distinguishes our technology from that of other domestic peers. In other words, EVsion’s AI not only assists physicians in making diagnoses but also clearly communicates the basis and source of those diagnostic decisions.”

EVsion’s decision to enter the AI-based fundus imaging sector at this juncture was driven by more than just market trends. Ke Xin believes that market education has largely been completed, and the ophthalmology + AI market is gradually opening up, revealing increasing pain points and demands while also raising higher expectations for technology and talent—presenting a significant opportunity. However, if products fail to deliver practical value and become trapped in homogeneous competition, addressing industry needs will remain unattainable. In light of this, leveraging the founding team’s years of accumulated expertise, EVsion has proactively ventured into areas currently untapped by the industry. By avoiding the red ocean of diabetic retinopathy screening, the company has taken a distinctive approach to deeply analyze fundus images and extract richer quantitative data, thereby enabling multi-disease analysis of the fundus. This strategy provides robust data support for the future advancement of precision medicine and personalized healthcare.

“I come from an AI background. Previously, I was responsible for R&D related to fundus imaging at Daheng, and I have been working in image recognition for nearly 10 years. I am among the earliest pioneers in ophthalmic AI in China. A key characteristic of our team is our profound medical expertise, rather than a pure IT background. The founding team consists of experts in the field of medical imaging, which ensures our unique advantages in technology development and enables our products to achieve clinical implementation more rapidly,” said Ke Xin.

Ke Xin further stated, “Currently, the general understanding of AI-based medical imaging remains somewhat superficial. Achieving genuine clinical implementation in this industry entails significant barriers. Technology, data, medical expertise, and practical application scenarios all constitute industry hurdles; however, the most critical barriers are technology and talent. Artificial intelligence is not merely about acquiring images for training purposes; a full-stack AI approach must be adopted to avoid deficiencies in addressing real-world challenges. As widely recognized, the algorithm engineers who developed AlphaGo were themselves expert Go players. Similarly, for medical AI to be successfully implemented, technical personnel must possess a deeper understanding of medical imaging. Our product logic involves algorithm engineers leveraging computer vision to first extract key elements from fundus images, followed by disease-specific analysis. Consequently, our results are intuitive, interpretable, and quantifiable. This ensures that the information derived by AI is transparent to both physicians and patients—a capability where domestic peers currently share similar technological proficiency.”

At the end of the interview, Ke Xin also put forward a viewpoint that is being continuously validated: “Currently, many medical AI practitioners have limited medical backgrounds, which has led to the biggest challenge in the industry today—difficulty in product implementation. In healthcare, we should maintain a sense of reverence, ensuring that products truly solve problems and assist doctors rather than replace them. Doctors cannot be replaced; however, with AI, they can better focus on their core responsibilities. This is the true progress of our era!”