AI in Hospitals Proven Effective for Doctors After Four Years: Synyi Intelligence Submits IPO Prospectus

Man vs. Machine? Yes, since the concept of “AI + Healthcare” emerged, you’ve likely heard this term countless times. Much like concert tours, companies stage showy AI-versus-doctor competitions every year. As a result, doctors have grown resentful, enterprises have become weary, and the industry has largely returned to its usual divisions.

However, a recent initiative by a determined AI company, in collaboration with the Shanghai Children’s Medical Center (one of the primary institutions comprising the National Children’s Medical Center), has experimentally demonstrated that the combination of artificial intelligence and physicians is indeed superior to manual data extraction by physicians alone in scientific research data collection. Their objective was not to stage a publicity stunt or proclaim AI’s superiority over humans, but rather to employ rigorous randomized controlled trial (RCT) methodologies to convey to clinicians that they may genuinely benefit from such an AI product, as it makes research work as convenient and user-friendly as using an electric toothbrush.

The development of natural language processing (NLP) technology has provided new technical avenues for physicians to better utilize unstructured electronic medical record (EMR) data. In recent years, the healthcare industry has increasingly recognized the use of NLP technology to extract valuable information from clinical texts, enabling the identification of patients who meet inclusion criteria or facilitating large-scale data extraction and analysis. Surprisingly, although NLP has been employed in hospitals in China for over four years, no institution had previously applied rigorous methodologies and scientific data to substantiate this practice. This study precisely fills that gap.

Today, they published their research findings in JMIR Medical Informatics, an international SCI-indexed journal specializing in medical informatics (2019;7(3):e13331; Impact Factor: 3.188). Following in-depth discussions with the head of technology at SYNYI·AI, VCBeat not only summarized the core content of the paper but also further explored the value of natural language processing (NLP) beyond clinical data entry.

This study engaged methodological experts in clinical research to participate in the study design, and invited internationally renowned authorities in the field of medical informatics to contribute to the preliminary design. A randomized, double-blind, controlled clinical trial was conducted, enrolling 24 volunteers with a background in clinical medicine as research participants.

As the technical support provider for the trial, the medical AI company SYNYI·AI specially developed an information extraction system based on the i2b2 standard (NLP-MIES) for volunteers. With this system, doctors no longer need to engage in tedious manual data entry; instead, they only need to confirm and correct pre-set information, thereby eliminating the time required for sequential data input.

Workflow of the Natural Language Processing System for Medical Information Extraction

To ensure the scientific rigor and integrity of the trial, Shanghai Children’s Medical Center will provide standardized training to all volunteers prior to the commencement of the study, ensuring that they are proficient in using the NLP-MIES system and the eCRF (electronic Case Report Form) application for data entry. After signing the written informed consent form, volunteers will be randomly assigned to either the manual group or the NLP-MIES-supported group, thereby officially initiating the trial process.

Randomized, Double-Blind, Controlled Clinical Trial Site

Additionally, each volunteer was provided with a laptop and required to complete all data entry within the specified timeframe of the trial. Through these design features, the trial fully simulated a real-world electronic Case Report Form (eCRF) completion process, thereby ensuring that the trial could demonstrate that NLP technology can enhance the efficiency of scientific research data collection. The trial results are shown in the table below:

The results showed that it took a physician an average of 73.28–103.79 seconds to manually enter a complete medical record, whereas the AI-assisted approach required only 49.42–69.73 seconds. In terms of accuracy, manual entry alone achieved 84.21–90.42%, fully automated AI achieved 84.15–92.77%, and AI combined with human verification reached 92.19–97.17%. Compared with the manual group, the AI group improved accuracy by 15% for congenital heart disease and 18% for pneumonia, while reducing time consumption by 31% and 33%, respectively. In short: Using AI for data collection enables physicians to work faster and more accurately!

The scope of scenarios in which a technology can be applied is often related to its level of maturity, representing a process of quantitative change leading to qualitative transformation.

Take the currently popular 5G technology as an example. It is precisely the leapforward development in network speed, driven by Huawei’s mature 5G technology, that has activated many medical scenario-based applications that were previously difficult to implement or offered poor user experiences. Examples include wireless monitoring, real-time remote consultations, wireless surgical demonstrations, and remote robotic ultrasound and endoscopic examinations.

Another example is computer vision in the field of artificial intelligence. Once facial recognition technology achieved a breakthrough, AI became capable of identifying human movements in videos and subsequently predicting various action trajectories. Alibaba leveraged this technology to enable facial recognition payment, giving rise to emerging physical entities such as unstaffed, self-service supermarkets and restaurants. These application scenarios were unimaginable before the breakthrough in facial recognition technology.

Today, SYNYI·AI’s NLP technology has been refined over three years. It can interpret medical records with the proficiency of a physician while efficiently structuring textual data such as medical literature and patient records, thereby meeting diverse demands beyond scientific research. Once hospitals gain the capability to parse their textual data, they can activate various analytical models and accomplish many tasks that were previously unattainable with conventional information systems.

In this regard, Liu Shijian, Director of the Biobank at Shanghai Children’s Medical Center and corresponding author of the paper, also stated that the application of NLP extends beyond scientific research. Extracting data is merely the first step; physicians place greater emphasis on the utilization of clinical data. A prime example is the AI-based Clinical Decision Support System (CDSS) used in the electronic medical record grading system of the National Health Commission. Such systems can assist clinicians in making clinical decisions and help hospitals standardize clinical pathways, offering broad prospects for future application.

Perhaps, as Zhang Shaodian, CEO of SYNYI·AI, stated: “The hallmark of a technology’s success is that you use it without being consciously aware of its presence—much like spring rain, which nourishes all things silently and unobtrusively, arousing no surprise. Artificial intelligence can similarly be embedded within physicians’ workflows, much like internet technology, truly becoming a partner to doctors.”

In future healthcare scenarios, AI will ultimately be ubiquitous.

To download the original paper, pleaseClick here, Access code: aw5p