Dr. Song Shijing of Capital Medical University: Integrating Clinical Practice and Research, Cross-disciplinary Collaboration Delivers Synergistic Impact

Before the formal interview, Song Shijing shared some good news with me—he has recently joined a postdoctoral research station while serving as an attending physician at Beijing Obstetrics and Gynecology Hospital.

This news piqued my curiosity: As a physician deeply rooted in clinical practice, how exactly does one balance scientific research with daily work?

In fact, he did not provide a specific answer to this question during the interview. In Song Shijing’s view, there is no such thing as “balancing” scientific research and clinical work. Unresolved issues in clinical practice guide the direction of research, while research findings ultimately return to clinical application. The two are complementary and constitute an organic whole.

Guided by this mindset, Song Shijing consistently proactively identifies unresolved clinical pain points in her daily work and meticulously documents and organizes them.

Through continuous documentation and reflection, his team discovered that applying artificial intelligence technology to prenatal ultrasound examinations can significantly improve the detection of early fetal chromosomal abnormalities, thereby helping to prevent and control birth defects in newborns to a certain extent.

This represents a vast blue-ocean market. On one hand, China is among the countries with a high prevalence of neonatal birth defects. According to estimates in the “Report on Birth Defect Prevention and Control in China,” the incidence rate of birth defects in China is approximately 5.6%, close to the 5.57% observed in middle-income countries worldwide. On the other hand, integrating artificial intelligence into medical imaging is an emerging sector widely favored by investors, holding significant potential for future development.

Following a series of preliminary research and preparatory work,His team has anchored its focus on AI-driven diagnosis for early fetal ultrasound imaging.。

Currently, theyFor the first time, a radiomics-based prediction model for chromosomal abnormalities using fetal ultrasound imaging features has been developed. This model employs a non-invasive approach for early fetal screening, assisting clinicians in making more accurate diagnostic judgments and treatment decisions regarding chromosomal abnormalities.

Scientific research has never been a “solo endeavor”; it invariably requires team members with complementary skills to jointly explore and advance knowledge.

The composition of Song Shijing’s team is highly diverse, including not only Chief Physicians with profound expertise in the prevention and control of neonatal birth defects—such as Dr. Sun Yongqing, a postdoctoral fellow in medicine, and Dr. Feng Li—but also Researcher Dong Di, who specializes in machine learning and artificial intelligence, along with Research Secretary Wang Jingjing.

So, how exactly is such a multidisciplinary team from diverse fields assembled?

In an interview, Song Shijing stated, “Although we come from different fields and possess distinct professional expertise, the integration of these capabilities creates a remarkable synergy that enables us to address a common challenge: early fetal screening. Thus, despite our diverse backgrounds, we share a unified goal. It is precisely this shared objective that has unconsciously coalesced us into a cohesive team.”

Song Shijing is a prime example. As a physician, he deeply recognizes that prenatal ultrasound examination is an exceptionally challenging technique, requiring operations through the maternal abdominal wall and encountering numerous obstacles during fetal assessment. “Many anatomical structures and functional aspects need to be evaluated. I aim to minimize these impediments and enhance the accuracy and precision of diagnosis.” This has been Dr. Song’s aspiration since he became deeply involved in clinical practice, as well as the focus of his ongoing research.

However, overcoming the challenges of early screening is no simple task; every step forward encounters a variety of obstacles. For Song Shijing, it was fortunate to have the guidance of Professor Wu Qingqing, which enabled the technological research and development to advance and ultimately achieve today’s results.

Professor Wu Qingqing is a pioneer in the Department of Ultrasound at Beijing Obstetrics and Gynecology Hospital. As the Vice President of the hospital and the primary leader in maternal and child healthcare, her dedication to her profession is evident to all. “Professor Wu Qingqing often says that we are the first to witness a baby’s heartbeat, and we must take responsibility for this profession and for new life,” said Song Shijing.

Professor Wu Qingqing not only cultivated Song Shijing’s spirit of inquiry and sense of responsibility in the field of neonatal early screening, but also provided many constructive suggestions during the actual research and development process, helping him avoid many detours. Meanwhile, mutual learning and collaboration within the team contributed significantly to his professional growth.

Currently, artificial intelligence technology is permeating every sector of the healthcare industry, andMedical imaging is the area where artificial intelligence has been most rapidly implemented in the healthcare sector.。

In current clinical diagnosis, a large volume of medical images requires processing and interpretation. However, traditional practices rely on manual analysis performed exclusively by experienced physicians. This workflow is not only time-consuming but also generally yields suboptimal diagnostic accuracy.

The integration of artificial intelligence can effectively address the aforementioned limitations. AI not only stores vast amounts of information but also continuously refines its algorithms through learning, thereby ensuring the accuracy and stability of medical image analysis.

This holds significant implications for the early screening of birth defects. Song Shijing stated that in medical imaging, artificial intelligence has achieved high accuracy in diagnosing CT and MRI scans; however, its application in the field of ultrasound remains limited. This is primarily due to inadequate standardization and quality control of ultrasound images, which results in poor analyzability of the imaging data.

To address this pain point, Song Shijing’s team proposed numerous solutions. “Our teamStandardized control of fetal imaging quality is excellent., laying a solid foundation for the integration of artificial intelligence. Building on this foundation, machine learning (the cornerstone of AI) is employed to analyze various fetal parameters, which holds significant value for the diagnosis of fetal chromosomal abnormalities. Subsequently, artificial intelligence—primarily deep learning techniques for medical imaging—is applied to image analysis to validate its clinical utility.

Currently, the ultrasound imaging technology developed by Song Shijing’s team enables visualization of facial, cervical, and intracranial structural parameters in fetuses during the first trimester (11–13 weeks of gestation), thereby facilitating prediction of chromosomal abnormalities.This detection model has been extensively tested at Capital Medical University, demonstrating favorable clinical outcomes.

So, can the detection model still deliver its intended benefits when applied to different hospitals, ethnic groups, and populations with distinct lifestyle habits?

No definitive conclusion has been reached yet, but Song Shijing’s team is already working in this direction and has listed it as a goal for the next phase.“We are currently conducting joint model validation with institutions in East and Southwest China, and will carry out multi-center trials and assessments across broader regions and ethnic groups in the future, aiming to obtain more scientific and precise results.”

In Song Shijing’s view, only through continuous model validation and the collection of patient data for deep learning can the precision of the model be improved; likewise, only by continuously enhancing model precision can the efficiency and accuracy of diagnosis be ensured. This is a mutually reinforcing process.

In addition to validating the current diagnostic models, Song Shijing also mentioned that predictive models will continue to be developed in the future.“For instance, when certain diseases have just been identified and their impacts on children and pregnant women remain unclear, predictive models can effectively address this uncertainty.”

Drawing on his own experience, Song Shijing candidly noted that researchers tend to have a relatively focused mindset, prioritizing how to advance the technology and refine the data and types used in models. However, to truly implement the technology and develop marketable products, one must incorporate a business operations mindset—a divergent way of thinking that requires a certain level of understanding in fields such as law and finance.

However, market-oriented operational thinking is precisely an area where the vast majority of scientific researchers lack proficiency, and Song Shijing is no exception. When discussing the pain points in translating research achievements into practical applications, he frankly acknowledged that his team’s approach to commercial promotion was not yet mature. How can this issue be addressed? Song Shijing has placed his hopes on professional market-operation agencies.

However, identifying suitable operational partners and establishing robust collaborative relationships with them presents a significant challenge for researchers. Any misstep could delay the translation of research outcomes; in severe cases, it may result in total losses of both personnel and financial resources, where the project fails to be commercialized and the researchers themselves become entrapped.

Therefore, in Song Shijing’s view, when collaborating with external market institutions, it is essential to first clarify one’s own needs and identify what the current research project requires most. Secondly, potential partner institutions should be evaluated from a multidimensional perspective: whether they recognize you and your technology, and whether their core competencies can effectively compensate for your weaknesses.

Finally, it is essential to have a plan and maintain a long-term perspective. The translation of scientific and technological achievements is a protracted process. As researchers, we must make thorough plans at the outset and, during implementation, remain vigilant against being blinded by short-term gains, thereby sacrificing greater interests for minor ones.

Therefore, the team to which Song Shijing belongs participated in2022 The 2nd Biomedical High-Value Patent Project Selection & National Conference on Medical-Engineering Integration and New Medical Technologies / Sci-Tech China Industry Matchmaking Meeting, with the hope of leveraging this platform to bring together teams from diverse scientific disciplines and industries to engage in ultrasound imaging research, thereby attracting broader societal attention—including enterprises, hospitals, and families expecting children—to early fetal chromosomal screening.

“University researchers with ideas for translating their findings into practical applications should actively engage in exchanges and collaborations with specialized teams and project groups. Only by partnering with professional technology transfer organizations can they achieve a synergistic effect where 1+1>2. Ultimately, it is most appropriate to leave specialized tasks to specialized professionals,” said Song Shijing at the end of the interview.