Infervision Files IPO Prospectus: China's First Deep Medical AI Company with Clinical Deployment

Recently, tech giants such as Google, Facebook, Amazon, IBM, and Microsoft have been investing heavily in artificial intelligence, with its development momentum being astonishing. However, some AI-assisted diagnostic devices purchased by domestic hospitals from abroad are left idle due to low accuracy. Overseas equipment uses data from foreign patients during the data collection process, which may not necessarily be suitable for use with domestic patients and hospitals.

Founded in early 2015, Infervision is the first deep medical AI company in China to achieve clinical application. It aims to leverage artificial intelligence technology and domestic data to introduce AI into the medical field. The company seeks to use AI to replace physicians’ burdensome and repetitive tasks, assist in medical diagnosis, alleviate shortages in medical capacity, and liberate high-quality medical resources, thereby making affordable, high-quality healthcare accessible to every household. VCBeat (WeChat ID: vcbeat) conducted an in-depth interview with Chen Kuan, founder of Infervision, and published a detailed report on him.

More than 50% of radiologists work over eight hours a day, with 20.6% averaging more than ten hours daily. Many physicians report that “radiation leave” exists in name only. Numerous doctors have expressed their hope to one day enjoy the long-denied radiation leave and public holidays, allowing them to spend more time with their families.

According to data from VCBeat’s Eggshell Research Institute, the annual growth rate of medical imaging data in China is currently approximately 30%, whereas the annual growth rate of radiologists is only about 4.1%, resulting in a gap of 23.9%. The increase in the number of radiologists falls far short of the surge in imaging data. This implies that radiologists will face escalating pressure in processing imaging data in the future, potentially exceeding their capacity by a wide margin.

Radiologists face a heavy daily workload, particularly between 3:00 and 4:00 PM, when fatigue peaks and concentration wanes. This increases the likelihood of overlooking imaging findings and significantly raises the risk of diagnostic errors. Consequently, physicians seek to reduce work-related stress and enhance efficiency, while hospital administration aims to strengthen oversight of medical service quality.

Chen Kuan graduated from the University of Chicago two years ago with a master’s degree in mathematics and economics. He is currently pursuing dual doctoral degrees in finance and economics at the same university, under the supervision of two Nobel Laureates in Economics, James Heckman and Lars Hansen. Chen began working with artificial intelligence technologies in 2011.

In 2012, he and his friend built a model to predict the outcome of the presidential election between Obama and Romney based on status updates posted by Twitter users. Many media outlets approached them, eager to purchase their findings, which made him realize the vast market potential of artificial intelligence.

However, after returning to China in 2014, Chen Kuan found the gap between the national conditions of China and the United States to be substantial, with artificial intelligence not yet receiving widespread attention domestically. He was also uncertain about which industries could benefit from his technology. During this period, Chen Kuan gradually began engaging with the healthcare sector, where he discovered that physicians specializing in medical imaging diagnosis faced enormous workloads and were prone to errors due to fatigue. Moreover, radiologists were scarce, particularly in remote areas.

After gaining an in-depth understanding of the healthcare industry, Chen Kuan also recognized the shortage of high-quality medical resources at the primary care level and the waste of resources in diagnosing common diseases at large hospitals. Consequently, he ultimately chose to pursue a career in the healthcare sector.

Physicians’ diagnoses are not based solely on imaging studies; they also incorporate patients’ health information, medical records, and laboratory test results. For artificial intelligence to assist physicians in auxiliary diagnosis, it must comprehensively integrate multifaceted information. In terms of data acquisition, Infervision has partnered with multiple Grade A tertiary hospitals to leverage their clinical data, striving to ensure comprehensive data coverage and broad applicability to the entire patient population in China.

Given the abundance of data on thoracic diseases, Infervision started with this area to first develop a product ready for clinical application. After accumulating experience, it then expanded into other disease areas.

Infervision ultimately aims to enable its products to assist physicians in diagnosis and automatically generate diagnostic reports, thereby improving diagnostic accuracy while reducing physicians’ workload. To this end, the company has been refining its algorithmic models while collaborating closely with clinicians to produce preliminary diagnostic reports; if physicians are not satisfied, the reports are revised accordingly based on their feedback.

Currently, the automated diagnostic reports generated by their products in the cardiopulmonary field exhibit approximately 90% similarity to physicians’ reports. Moreover, Infervision can now iterate its products roughly every three days, achieving progressively higher accuracy. Chen Kuan told reporters that this constitutes Infervision’s current competitive advantage.

As algorithmic models continue to be refined and iteration accelerates, service offerings are gradually expanding into areas such as head, abdomen, femoral head, pathology, and ultrasound imaging, covering more than 100 diseases. Leveraging experience in thoracic diseases, research progress in other domains will be significantly accelerated.

Currently, Infervision's products are mainly applied in four major sectors: hospitals, traditional medical IT giants, health examination centers, and internet healthcare.

Hospital: Some top-tier Grade A tertiary hospitals have begun adopting Infervision’s deep artificial intelligence system, integrating its AI alert system into physicians’ information portals. Physicians need only upload patients’ medical images; the system then generates a diagnosis and produces an initial diagnostic report, which physicians can review and supplement as needed. This has significantly improved physicians’ work efficiency.

Traditional Medical IT Giants: Infervision has signed OEM and strategic cooperation agreements with multiple international medical IT giants, integrating its AI systems into their comprehensive solutions to jointly undertake hospital digitalization projects.

Health Examination Center: The health checkup industry has also begun to adopt Infervision’s intelligent screening services, leveraging this technology to enhance examination efficiency and improve service quality.

Internet Healthcare: Due to the shortage of radiologists in remote areas, telemedicine serves as an effective solution. By integrating Infervision’s intelligent screening system for preliminary assessments, it can reduce the burden on patients traveling for medical care and alleviate the difficulty in accessing physician resources in underserved regions.

By the end of 2016, Infervision will continue to refine the clinical applications of its products for thoracic diseases. Starting in 2017, it will launch nationwide marketing campaigns across China while expanding into other disease areas.

Infervision currently has more than 20 employees, most of whom are technical professionals. In February 2016, the company completed its RMB 12.5 million angel financing round, with investors including Inno Angel Fund, Zhenyun Venture Capital, and Lv Chuanwei, the former CEO of Kuaidi Dache.

This September, Infervision won the championship at NVIDIA’s GTC (GPU Technology Conference) in China. The GTC attracts developers, researchers, and technical experts from leading companies, universities, research institutions, and government agencies across China and around the world, showcasing the most significant innovations in GPU technology across various industries, including deep learning, supercomputing, virtual reality, artificial intelligence, and intelligent robotics. Infervision will compete with top artificial intelligence companies at the global GTC conference in Silicon Valley, USA, next year.