Huiying Medical's Wang Jie: The Value of AI in Healthcare Lies in Extending Services Downstream

In April 2015, at a time when artificial intelligence technology remained largely peripheral to healthcare, Dr. Chai Xiangfei, then a postdoctoral fellow at the Stanford Cancer Center in the United States, decided to return to China and founded HY Medical, initiating the integration of AI into the medical field. At this juncture, HY Medical held a first-mover advantage.

The average three-month product iteration cycle in the field of artificial intelligence has plunged numerous startups into fierce competition for market share. Over the past four years, HY Medical has completed its evolution from digitization to mobilization and intelligentization of medical imaging. Its AI products and cloud platform have been gradually implemented in more than 800 hospitals, with digitization and mobilization forming the solid competitive barriers that define HY Medical today.

At the “2018 Future Healthcare Power Forum” hosted by VCBeat, Wang Jie, Vice President of HY Medical, outlined the trends in AI development for 2018 and summarized HY Medical’s growth trajectory, aiming to provide insights for like-minded professionals. VCBeat has compiled his views as follows.

Wang Jie, Vice President of HY Medical

Modern medicine is advancing at an uneven pace. This imbalance is evident not only in macro-level aspects such as geography, human resources, and technology, but also within the field of medical imaging, where supply and demand are out of balance even when considering solely the diagnostic phase. Wang Jie believes that the factors driving the rapid growth of imaging data mainly include the following two aspects:

First, while advances in medicine have yielded new therapies for many chronic diseases, the incidence of certain chronic conditions has risen rather than declined amid accelerating population aging and an upgrading consumption structure. Furthermore, growing health awareness among the public has led to a sharp increase in demand for medical services.

Second, the unequal distribution of medical resources on a macro level has further exacerbated the aforementioned issues. Patients from less developed provinces and municipalities flow into more developed regions, while the geographic distribution of medical schools and the quality of their student intake over the past decade have not seen substantial improvement. First-tier cities tend to concentrate a large number of top-tier hospitals and prestigious medical schools; talent from these elite institutions feeds back into the hospitals, further intensifying the agglomeration of medical resources and creating a siphon effect. Furthermore, constrained by structural reforms on the supply side of society, the prolonged cycle of professional development, and high labor intensity, the overall supply of medical resources is showing a downward trend. Against this backdrop, hospitals need to explore alternative avenues to enhance resource supply.

Both of the above issues can potentially be resolved through AI. Ultimately, AI will serve to augment capacity in the healthcare sector: supplementing labor at top-tier hospitals and enhancing technical capabilities at second-tier institutions. This augmentation will drive disruptive transformation in healthcare, gradually shifting medicine from an experience-based discipline to a data-driven one, thereby advancing precision medicine.

The development of AI is by no means an overnight achievement; throughout 2018, AI subtly transformed the healthcare sector.

In January, Andrew Ng’s research team at Stanford University open-sourced MURA, a dataset comprising 40,000 X-ray images of the human upper extremities, and utilized this dataset to train convolutional neural networks (CNNs) to detect and localize abnormalities in the X-rays. According to the study, over 1.7 billion people worldwide currently suffer from musculoskeletal disorders, resulting in approximately 30 million emergency department visits annually. The open-source dataset will facilitate deeper engagement in medical AI research by a broader community of AI practitioners.

In February, *Cell* featured an AI tool developed by a Chinese team, an artificial intelligence system capable of accurately diagnosing two major categories of diseases: ocular disorders and pneumonia. The tool effectively classifies images into cases of macular degeneration and diabetic retinopathy. In contemporary medical research, there is a growing acceptance among physicians, scholars, and professional journals of using AI modeling to replace the traditionally time-consuming in vivo biological experiments, marking a highly significant shift.

In April, HY Medical, in collaboration with the Department of Vascular Surgery at the Chinese PLA General Hospital, launched the “AORTIST 2.0: AI Cloud Platform for Aortic Research.” This platform features the world’s first AI-based automatic segmentation method for Type B aortic dissection, addressing three core challenges in the surgical management of this condition: precise measurement, prognosis prediction, and remote follow-up. Under normal circumstances, aortic imaging can comprise up to 1,000 slices, requiring physicians to spend roughly an afternoon analyzing the data. In contrast, this AI solution completes image analysis within 20 minutes through scanning and reconstruction, significantly enhancing hospital efficiency.

In May, a study published in ANNALS OF ONCOLOGY described the development of a deep learning convolutional neural network (CNN) trained on more than 100,000 images of malignant melanomas and benign nevi to identify skin cancer. This marked the first time scientists demonstrated that a CNN, as a form of artificial intelligence or machine learning, could diagnose skin cancer more accurately than experienced dermatologists.

In September, Professor Liao Hongen’s research group, comprising distinguished experts from the Department of Biomedical Engineering at Tsinghua University School of Medicine, leveraged artificial intelligence to analyze magnetic resonance imaging (MRI) features from a large cohort of patients with brainstem gliomas. By deeply mining the associations between these imaging features and specific genes, the study not only provided clinicians with genetic diagnostic evidence but also identified imaging and clinical parameters closely linked to genetic profiles, thereby enhancing physicians’ diagnostic expertise. The series of research findings were published in IEEE Transactions on Biomedical Engineering, a prestigious journal in the field of biomedical engineering.

Also in September, a team led by Dr. Eric Deutsch from France trained artificial intelligence using CT images from cancer patients, resulting in an AI platform capable of accurately predicting the therapeutic efficacy of PD-1 inhibitors based on patients’ CT scans. The median overall survival for patients predicted to respond favorably was 24.3 months, more than double that of patients predicted to be non-responders (11.5 months).

While the world in 2017 was still debating the “Frankenstein” question of “AI replacing doctors,” a concern that has haunted humanity for over two centuries, academic summits in 2018 had already begun discussing how to apply AI technology to the treatment of various diseases. Integrating AI into clinical scenarios has virtually become a consensus among the global academic, industrial, and even political communities.

As a participant in this trend, HY Medical is also continuously promoting the driving role of AI technology in medicine. Wang Jie gave a brief introduction to the development of HY Medical.

In the 1.0 era, HY Medical’s chest X-ray screening, pulmonary nodule screening, fracture detection, and stroke assessment primarily addressed auxiliary screening by detecting and localizing lesions, generating structured reports, and providing preliminary screening results for physicians’ reference.

We currently believe that AI is entering the 2.0 era, integrating into the entire healthcare journey of patients. The role of medical imaging will no longer be limited to screening and diagnosis; HY Medical can also incorporate patients’ clinical data, pathological data, examination data, genetic data, and follow-up data. By leveraging such multimodal data, it can assist in surgical guidance, medication management, and prognosis prediction.

In terms of business model, HY Medical aims to build a comprehensive, full-cycle data platform. By aligning with the needs of patients and hospitals, it seeks to refine an integrated product that covers the entire healthcare lifecycle and service chain, using the platform to feed back into and enhance its AI products, rather than merely employing AI for auxiliary screening.

In terms of data, HY Medical has engaged in research collaborations with more than 500 top-tier hospitals. These collaborations encompass radiomics and deep learning research, as well as the development of disease-specific case databases. Furthermore, HY Medical cleanses medical data based on existing healthcare information systems to establish digitized medical imaging infrastructure. This capability fundamentally distinguishes HY Medical from its competitors.

In terms of patient management, HY Medical provides digital imaging services to patients. Leveraging NLP technology, it also offers intelligent interpretation of imaging reports. Patients can directly access their medical imaging data and diagnostic results via WeChat, significantly enhancing their healthcare experience. Furthermore, on the HY Medical platform, patients who have received their imaging results can consult with imaging specialists online for further diagnosis.

Today, HY Medical has gathered top AI scientists and medical experts from around the world, with more than 50 professionals possessing backgrounds in both medicine and engineering. The company has established multiple AI laboratories in collaboration with prestigious institutions such as the Tsinghua Strait Research Institute, Stanford University, and Intel. With over 500 projects under development, the integrated closed-loop structure combining industry, academia, research, and application has begun to yield tangible results.

In terms of implementation, HY Medical’s products have been deployed in more than 800 hospitals, with over half of them being secondary or primary care institutions. Wang Jie stated, “At this stage, AI can deliver the greatest value at the grassroots level. In these hospitals, whether by reducing physicians’ workload, improving diagnostic accuracy, or facilitating tiered diagnosis and treatment, our technology has demonstrably enhanced services for both hospitals and patients.”

Looking ahead to 2019, data will once again become the focal point of competition; AI will cover more disease types, enter more departments, streamline more processes, and reach a stage of inclusive implementation. Meanwhile, these innovative technologies will serve as new drivers for healthcare reform by decentralizing resources, enhancing capacity, and advancing precision medicine.