Capturing the RMB 200-Billion Liver Cancer Screening Market: AI and NGS Await Their Breakthrough

Liver cancer mortality in China ranks second among all cancer-related deaths. The latest national cancer statistics, released in January 2019, show that malignant tumors account for 23.91% of all causes of death among residents. In 2018, there were 841,080 new cases of liver cancer worldwide annually, with 55% of these new cases occurring in China; however, only 10% of patients survive beyond five years.

However, if detected early and treated surgically, the 5-year survival rate for liver cancer can reach as high as 80%–90%. Early detection and early treatment have become a shared goal for both patients and physicians.。

Regarding the market size of liver cancer, in April 2018, Southwest Securities provided data indicating that there are over 400 million chronic hepatitis B and C infections globally, with approximately 1.4 million deaths annually due to hepatitis virus infections. Among liver cancer deaths, 80% have a background of hepatitis B or C infection.

Compared with individuals not infected with hepatitis viruses, patients infected with hepatitis B or C are more prone to developing liver cirrhosis. Patients with liver cirrhosis face a significant risk of hepatocellular carcinoma, with an annual incidence rate as high as 4–6%. There are nearly 10 million patients with liver cirrhosis in China.For individuals at high risk of liver cancer, assuming one screening per person per quarter at a cost of RMB 5,000 per test, the market size for early diagnosis of liver cancer would reach as high as RMB 200 billion.。

Diagnostic tools for liver cancer are not lacking; hematological and imaging tests have become widely available in primary care settings and health checkup centers. So why is it that, despite the widespread adoption of liver cancer screening at the grassroots level, most patients are still diagnosed at intermediate or advanced stages? What are the limitations of current screening methods, and how can AI and NGS technologies help companies compete in the billion-dollar liver cancer screening market?, VCBeat seeks to find the answer.

Guo Pengfei, Executive Deputy Director of the Southeast Institute for Liver and Gallbladder Health Big Data at Mengchao Hepatobiliary Hospital, told VCBeat thatCurrently, liver cancer screening methods are divided into two major categories: one is hematological testing, and the other is imaging examination.。

Hematological screening primarily refers to the detection of alpha-fetoprotein (AFP). AFP-based screening relies on reagent kits. Although this method is convenient to perform, both domestic and international studies have demonstrated that alpha-fetoprotein, the current serological diagnostic marker for liver cancer,Specificity and sensitivity remain insufficient for the diagnosis and early warning of very early-stage hepatocellular carcinoma, with a discrimination rate of only approximately 60%.。

To improve the accuracy of screening, clinical practice now involves testing multiple hematological markers based on specific needs, typically including alpha-fetoprotein (AFP), des-gamma-carboxy prothrombin (DCP), and AFP-L3. Simultaneous detection of these three markers can increase the early detection rate of liver cancer to approximately 90%.

Integrated into clinicians' workflow, B-mode ultrasound is also a common screening modality for liver cancer. Other frequently used imaging examinations include computed tomography (CT) and magnetic resonance imaging (MRI).

Guo Pengfei stated,When the general public undergoes physical examinations or targeted liver cancer screening, the initial step typically involves combined alpha-fetoprotein (AFP) testing and ultrasonography. The approach of jointly assessing AFP, des-gamma-carboxy prothrombin (DCP), and Lens culinaris agglutinin-reactive fraction of AFP (AFP-L3) has not yet been widely adopted.。

Step 1: If an abnormal mass is detected during the initial examination, further confirmation will be performed using contrast-enhanced ultrasound (CEUS), contrast-enhanced CT, or MRI. Primovist-enhanced MRI in the hepatobiliary phase can even detect liver cancers smaller than 1 cm; however, the cost is relatively high, typically around RMB 1,200.

"In recent years, research on liver cancer screening using gene sequencing, high-throughput sequencing, molecular imaging, and other methods has gradually been carried out, but it has not yet formed large-scale applications in clinical practice."

Since liver cancer screening methods have become widely available, why are most patients still diagnosed at intermediate or advanced stages? In response, Guo Pengfei stated that the effectiveness of screening primarily depends on two factors:

First,Hepatitis patients have relatively weak awareness of the need for regular check-ups and of liver cancer.. These two factors lead to an excessively long turnaround time in the diagnostic process for liver cancer. After abnormalities are detected during screening at primary care facilities, patients need to visit tertiary hospitals for confirmation.However, screeners have insufficient awareness of liver cancer, failing to pursue timely confirmatory diagnosis and thereby missing the optimal window for intervention.。

The liver has a rich blood supply and active metabolism, so liver tumors grow rapidly, generally progressing from early to advanced stages within six months.

Additionally,The country still lacks a comprehensive medical information system. After screening detects abnormalities, many individuals are unsure which hospital to visit for further diagnosis and treatment, and do not know which hospitals have reputable hepatology departments. The country currently has no information guidance system for such needs.

Insufficient patient awareness of medical information further delays the diagnosis of liver cancer, resulting in most cases being diagnosed at intermediate or advanced stages.

Guo Pengfei stated that if patients present with suspected lesions, promptly seeking care at a high-level specialized liver hospital can facilitate rapid diagnosis and treatment. He cited a case where a patient completed the entire diagnostic and surgical process within five days at Mengchao Hepatobiliary Hospital.

However, in reality, it often takes patients 2–3 months from the detection of a suspected lesion to a confirmed diagnosis, causing them to miss the optimal treatment window. Therefore, facilitating rapid access for suspected cases to tertiary hospitals is critically important.

Second,Issues Regarding the Experience and Competency of Primary Care Physicians and First-Visit Clinicians. In screening programs, two key metrics are primarily monitored: the false positive rate (referring to cases identified as liver cancer that are not actually liver cancer, with lower values being preferable) and the missed diagnosis rate (referring to cases identified as non-liver cancer that are actually liver cancer, with lower values being preferable).

A high false-positive rate increases the subsequent workload for physicians, whereas missed diagnoses cause patients to lose critical time for seeking medical care and can lead to medical disputes. Therefore, missed diagnoses are a significant concern for physicians and should be avoided in clinical practice.

However, in actual clinical practice, missed diagnoses are often unavoidable.

As mentioned above, the first-line screening methods for liver cancer are alpha-fetoprotein testing and ultrasonography. The ability of ultrasonography to detect small hepatocellular carcinomas relies heavily on the operator’s extensive experience and advanced ultrasound equipment.

Many primary care physicians indeed lack sufficient proficiency in interpreting ultrasound images. Furthermore, the liver’s large size makes it prone to incomplete examination, while confounding factors such as cirrhotic nodules and benign hepatic nodules further complicate diagnosis. These issues have led to persistent problems of missed and misdiagnoses in B-mode ultrasonography.

Public awareness and knowledge of liver cancer screening are insufficient, requiring long-term cultivation and cannot be resolved in the short term. However, with the deepening integration of AI and healthcare, it is feasible to enhance the capabilities of primary care physicians through AI-assisted diagnostic technologies.

As one of the diagnostic modalities for liver cancer, computed tomography (CT) is not as widely used in primary screening as ultrasound. However, when patients present to tertiary hospitals for definitive diagnosis, physicians often recommend CT examination first. The clinical guidelines released in 2017 also explicitly state that physicians can use CT imaging or Primovist (gadoxetic acid) for the definitive diagnosis of liver cancer.

Moreover, the diagnosis and treatment of any disease constitute a complete workflow. For instance, the screening and diagnosis of liver cancer involve imaging, blood tests, and other procedures, each of which is indispensable. Leveraging advanced technologies such as AI to optimize specific tasks within any step of this workflow is highly meaningful.

The topic for the big data track of the 2019 Digital China Innovation Competition Finals was “Big Data in Healthcare: AI-Based Diagnosis of Liver Cancer from Imaging.”

The Fuzhou University–LinkDoc Technology Team, a joint effort by Fuzhou University, LinkDoc Technology, and Fuzhou Yixing Big Data, won the championship and the Most Commercially Promising Award for their project “Lesion Identification in Hepatobiliary Organs,” which leverages the integration of CT imaging and AI, standing out among more than 1,300 competing teams.

The team has developed a clinically deployable AI-assisted diagnostic system for hepatocellular carcinoma, featuring functionalities such as liver contour segmentation, lesion detection and delineation, quantitative analysis of lesion count, patient-level benign-malignant classification, multi-phase image comparison for radiological review, and generation of structured diagnostic reports.

These features hold practical application value for primary healthcare institutions in conducting early screening and diagnosis of liver cancer, assisting clinicians and radiologists in improving diagnostic efficiency and capability, and helping patients reduce contrast agent usage and the number of contrast-enhanced CT scans.

Luo Ligang, Co-founder and CTO of LinkDoc Technology, told VCBeat that applying AI technology to an auxiliary diagnostic system for liver cancer was a natural progression for the company. LinkDoc’s lung cancer auxiliary diagnostic system has already been deployed in several major hospitals, where it assists physicians in diagnosing more than 1,000 patients daily and has cumulatively served 600,000 oncology patients.

After perfecting its lung cancer products, LinkDoc naturally set its sights on liver cancer, the second most prevalent type of cancer. LinkDoc Technology was able to successfully develop this AI-assisted diagnostic system for liver cancer thanks, on one hand, to its accumulated expertise in AI technology, and on the other, primarily to its deepened understanding of clinical diagnosis and treatment workflows for liver cancer gained through processing hundreds of thousands of liver cancer cases, as well as its accumulation of resources from clinical experts.

It is precisely due to a thorough understanding of the diagnosis and treatment workflow for liver cancer, along with the assistance and support of numerous liver cancer specialists, that this computer-aided diagnostic system for liver cancer closely aligns with the clinical workflow of physicians and incorporates comprehensive functional modules.

Luo Ligang also stated that the current liver cancer computer-aided diagnosis system has only been validated in competitions. To achieve real-world clinical application, further refinement and optimization through close collaboration with clinical practice are required, and LinkDoc Technology is already making relevant preparations.

Currently, some AI companies' systems perform well in major competitions or at select partner hospitals, but they are not being used by hospitals during large-scale expansion and are not in actual operation.

Luo Ligang stated that this is primarily because the system tends to overfit on the predefined training data, meaning it performs well in certain scenarios but often fails to meet clinical requirements in more complex real-world clinical settings, thereby being shelved.

LinkDoc Technology has not pursued the sheer number of hospital deployments; instead, it requires that its products be genuinely utilized in every partner hospital. To this end, LinkDoc dedicates 1–2 months to integration and fine-tuning at each newly deployed hospital. Only after calibration and verification that performance meets expectations does the system become fully available to clinicians, thereby ensuring operational stability and tangible clinical utility for physicians.

On the Challenges of Implementing AI-Assisted Diagnostic Systems in Real-World Clinical Settings,Guo Pengfei also shared his perspective, noting that a major issue with current medical AI products is their excessive focus on algorithms while neglecting physicians’ workflows. In reality, only products developed in integration with clinical workflows can effectively serve physicians and justify hospital payment.。

If the R&D team does not understand physicians’ workflows and merely masters AI algorithms,And while a certain amount of real-world simulated imaging data has been acquired, it is far from sufficient to apply the model to clinical practice after training.。

The reason for this statement is, on the one hand, that the volume of data used for training was insufficient.. Guo Pengfei stated that AI is widely and maturely applied in the fields of security and transportation because these two domains involve hundreds of millions of data points, whereas the volume of medical imaging data is significantly smaller.

For instance, in this competition, the organizing committee prepared more than 10,000 CT images. With each patient contributing 200–300 CT scans, these 10,000-plus images were derived from only several hundred patients. This dataset is roughly equivalent to the annual outpatient volume of a high-level Grade IIIA hospital.

On the other hand,because such systems are uninterpretable. Systems derived from current artificial intelligence algorithms undergo various computational processes, making it difficult to determine whether the features identified by the system are meaningful. As these identified features are unexplainable and lack a factual basis, physicians cannot incorporate them into their reports; instead, this increases their workload by requiring additional review of these features.

So, how can AI products be developed in integration with physicians’ workflows? Guo Pengfei offered his recommendations.

The R&D team can integrate authoritative international or domestic guidelines and standards for the diagnosis and treatment of liver cancer, which define the major and minor features of the disease., classify or grade patients based on the number of features contained in their liver cancer CT images, with different grades yielding different diagnostic results. During this process, enable the AI-assisted diagnostic system to master these primary and secondary features.

If an AI-assisted diagnostic system masters the features identifiable by general physicians, it reaches the level of a general physician; if it masters the features identifiable by specialists, or even more than those recognized by specialists, it reaches the level of a specialist.

Ultimately,An AI-assisted diagnostic system composed of models capable of identifying different features acts like a series of sieves, filtering and recognizing these features layer by layer.。

In AI algorithm models, feature recognition learning requires significantly less data, and the identified features are interpretable. With such a system for assistance, doctors benefit from guidance akin to having an expert by their side. If the system also aligns with physicians’ workflow habits, they will be more inclined to adopt it.

However, this R&D approach requires more time for refinement and necessitates the involvement of numerous expert physicians to identify various clinical practice guidelines and differentiate disease conditions.

Guo Pengfei emphasized that relying solely on a single device for screening makes it difficult to arrive at an accurate diagnosis. By integrating AI products into information systems and incorporating them into physicians’ existing workflows, blood tests can be combined with imaging studies to enable more accurate assessments. This approach also helps primary care physicians make precise diagnostic judgments.

According to VCBeat, in addition to the combination of CT and AI for liver cancer diagnosis, the integration of ultrasound and AI can also be applied to liver cancer diagnosis. Currently, Deshang Yunxing is conducting related research.

As mentioned above, the two current screening methods for liver cancer each have their own limitations.

In the earliest stages, tumors are undetectable by most imaging modalities, such as CT, ultrasound, PET-CT, and MRI, because these techniques can only identify masses with a diameter greater than 1–50 px.

Furthermore, both domestic and international studies have demonstrated that existing serological diagnostic markers for liver cancer, such as alpha-fetoprotein (AFP), still exhibit insufficient specificity and sensitivity in the diagnosis and early warning of very early-stage hepatocellular carcinoma, with an early differentiation rate of approximately 60%.

In addition to hematology and imaging,Tumor genomics research is also a critical approach for the early screening and diagnosis of liver cancer: By applying tumor molecular markers to detect genomic alterations in tumor cells—such as mutations, deletions, rearrangements, methylation, amplifications, and insertions—during the early stages (Stage I and Stage II) of the disease, it provides guidance for early diagnosis, prognosis, and the selection of therapeutic strategies.。

In April 2018, the National Multicenter, Prospective Cohort Study of 10,000 Participants for Screening Ultra-Early Warning Biomarkers of Hepatocellular Carcinoma, jointly initiated by the National Center for Liver Cancer/Changhai Hospital of Naval Medical University, Nanfang Hospital of Southern Medical University, and Berry Genomics’ subsidiary (hereinafter referred to as Huirui Gene), was officially launched at the National Center for Liver Cancer in Shanghai.

In March 2019, at the Fourth Sino-US Hospital Management Symposium and the First C3 Sino-US Health Conference, Dr. Zhou Daixing, CEO of Berry Genomics and Director of HuiRui Gene, announced a major breakthrough in the PreCar project for early screening of liver cancer. Data showed that in controlled trials involving patients with liver cirrhosis and liver cancer, the sensitivity of liver cancer detection exceeded 97% with 100% specificity.

Sensitivity, in layman's terms, refers to the proportion of individuals with the disease who can be correctly identified. Specificity, in layman's terms, refers to the proportion of individuals without the disease who can be correctly excluded.

This result marks the successful development of an NGS-based classification model for liver cirrhosis and hepatocellular carcinoma (HCC) within the early screening program for liver cancer. The model effectively distinguishes patients with liver cirrhosis from those with HCC by leveraging features derived from plasma biomarker sequencing, thereby accurately identifying cirrhotic patients who have undergone malignant transformation.

The entire project will enroll 10,000–15,000 high-risk individuals in a three-year prospective cohort study. Upon completion of the three-year trial, the company may apply for market approval and commercialization of the diagnostic reagent. However, favorable clinical data obtained prior to that milestone would allow promotion through genetic testing services at partner hospitals, with such testing services expected to launch as early as 2020.

He Rui Gene, a participant in this project, was established in August 2017 as a spin-off from Berry Genomics, specializing in tumor genetic diagnosis. With joint support from Berry Genomics and professional investment institutions, He Rui Gene possesses multifaceted advantages in “technology, distribution channels, and capital.” Upon regulatory approval and market launch of this project, its brand and channel strengths are poised to significantly improve the current landscape of early liver cancer screening.

In addition to Huirui Gene, in March 2019, the prestigious academic journal Proceedings of the National Academy of Sciences (PNAS) published research findings on early liver cancer screening among a prospective cohort of hepatitis B virus carriers, based on cell-free DNA and protein biomarkers. This study was conducted collaboratively by Beijing Genetron Health Gene Technology Co., Ltd., the National Cancer Center, and the Cancer Hospital of the Chinese Academy of Medical Sciences. According to reports, after rigorous clinical validation, these findings hold promise for application in early liver cancer screening.

In addition to research on NGS, numerous companies and institutions are also engaged in the development of liver cancer vaccines.

In 2017, a research team at the University of California, San Diego School of Medicine discovered that a commonly used synthetic double-stranded RNA (dsRNA) can enhance various anti-tumor immune functions while investigating the pathogenesis of liver cancer. They posited that a vaccine might exist that could provide a certain degree of preventive effect against cancer development in high-risk populations for liver cancer.

In July 2018, Conba replied to investors' inquiries about the progress of the company's AFP liver cancer vaccine on an interactive platform, stating that the AFP project had completed the construction of the vaccine drug, with one patent application accepted, and was currently in stages such as preclinical pharmaceutical research.

The enormous demand will inevitably accelerate market development and drive greater investment in research. Whether it is artificial intelligence (AI) addressing the limited technical proficiency of primary-care healthcare workers, or next-generation sequencing (NGS) enhancing the differential diagnostic accuracy of liver cancer detection, these technologies are competing in their own ways for a share of the hundred-billion-yuan liver cancer screening market. Current participants are poised to capture a portion of this lucrative market.