The rapid evolution of technology in the medical sector has heralded a new era, one characterized by innovation and the promise of improved healthcare deliveryA noteworthy instance of this transformative shift was observed when an artificial intelligence (AI) robot successfully conducted an astounding 668 experiments in merely eight days, leading to the synthesis of 668 distinct compounds and the development of a groundbreaking chemical catalystThis remarkable achievement highlights the profound potential of AI to reshape the landscape of healthcare.
As a key driver of a new technological revolution and industrial transformation, AI is deeply influencing the medical and health fieldsCurrently, AI and big data are increasingly integrated into various domains such as pharmaceuticals, clinical practices, new drug development, and health management and interventionThis growing synergy promises to revolutionize healthcare access and quality, redefining how patients interact with medical providers and manage their health.
In recognition of these dynamics, the 21st Century New Health Research Institute has introduced a thematic series titled "The AI Medical Wave." This initiative aims to explore the challenges and advancements within the "AI + Medical" sector, focusing on corporate development trends and regulatory considerations, ultimately striving for high-quality growth in China's innovative pharmaceutical industry driven by digital empowerment.
Key players in this field, such as DeepSeek, are setting the stage for a significant breakthrough in medical AI by 2025, prompting vigorous responses from the capital markets as wellSeveral healthcare companies are beginning to adopt the DeepSeek R1 model within their operationsFor instance, on February 7, Airdoc announced the successful integration of their independently developed medical model with DeepSeek's R1, yielding notable advancements in clinical diagnostic efficiency and accuracy, alongside enhanced personalized health management experiences.
Other companies, including Yidu Tech and Hengrui Medicine, are also exploring the adoption of the DeepSeek technology, demonstrating a collective movement toward harnessing AI capabilities within healthcare frameworks
Advertisements
In a related surge, BGI Genomics witnessed a stock increase of over 24% on February 6 and 7, signaling strong investor confidenceFollowing these developments, they revealed insights from a discussion with institutional investors regarding their ambition to leverage AI technology in advancing their business operations, particularly within genetic testingThey noted that their large model has entered clinical application, significantly improving the efficiency of identifying pathogenic variants and thereby reducing interpretation costs.
A leading Chinese AI medical enterprise remarked on the transformative impact of DeepSeek, likening its emergence to a leap from the Stone Age to the New Stone Age for AI large models—underscoring the significance of such technological evolution for the healthcare sector.
This technological advancement signals the beginning of an industry-wide reconstruction driven by AI innovationOne of the pressing challenges that DeepSeek addresses is the limitations encountered in the conventional development of medical AI, particularly concerning data collection and annotationTypically, these processes are labor-intensive and costly, requiring vast amounts of accurately labeled data to achieve optimal performanceFor many grassroots healthcare institutions, the financial burdens associated with sourcing and annotating immense datasets remain a prohibitive hurdle.
Moreover, studies indicate that building a high-quality medical AI model can involve expenditures reaching millions of dollars, which can pose a significant barrier for many smaller healthcare facilities and companiesHowever, thanks to DeepSeek's advanced reinforcement learning technology, this dilemma may soon be alleviatedAs per official disclosures, the model utilizes extensive reinforcement learning techniques during a post-training phase, vastly improving inferential capabilities with limited manual annotations.
The key to DeepSeek’s success lies in its ability to continuously optimize its model through interaction with the environment, significantly diminishing reliance on extensive annotated datasets
Advertisements
This technological breakthrough not only reduces training costs but also enhances the model's generalization and adaptability, promoting a more efficient integration of AI within healthcare.
A representative from the wearable tech sector remarked on DeepSeek R1's implications for cost and efficiency in model training, highlighting potential applications in health-related algorithms such as health warnings and digital prescriptionsAdditionally, the support for multi-platform deployments by DeepSeek paves the way for its widespread application in medical AI.
Research from various international teams, including that of Hong Kong’s Polytechnic University, has elucidated that well-crafted reinforcement learning methods can bestow even smaller models with formidable inferential capabilities, often in a more straightforward and economical manner than traditional methods would allow.
Furthermore, as indicated by analyst Liu Kai from Everbright Securities in a report published on February 6, third parties may distill smaller parameter versions from DeepSeek R1, allowing deployment in various devices including smartphones, laptops, and smart home productsThis optimization means that healthcare providers can utilize less expensive hardware for AI applications without being burdened by the high costs associated with advanced computing equipment.
An equally notable aspect of the DeepSeek R1 model is its open-source strategy, which permits global researchers and healthcare institutions to utilize and modify the model freelyThis open accessibility significantly lowers technological entry barriers, enabling grassroots medical institutions to adopt cutting-edge AI technology at lower costs, thereby accelerating the democratization of healthcare AI.
Li Yinghua, Vice President of Kingmed Diagnostics, expressed optimism regarding how DeepSeek would enhance industry-specific modelsTheir specialized model in the medical testing sector, he noted, has its unique foundational model, akin to Tencent's "Mixuan" or Huawei's "Pangu." The choice of open-source frameworks was driven by an awareness that it would afford them broader opportunities for future developments while preventing potential constraints associated with closed-source models that may not keep pace with industry advancements.
Another critical facet of addressing the issue of healthcare resource distribution in China is the urgent need to bolster the competency of primary care providers
Advertisements
The uneven allocation of medical resources has long been a hindrance to the enhancement of public health levelsAccording to the 2023 National Health and Family Planning Statistics, there are currently over 1,070,000 medical institutions nationwide, with grassroots healthcare facilities representing the majority, while hospitals account for a mere 3.5% yet conduct nearly half of all medical consultations.
The training of primary care general practitioners (GPs) is thus essential, ideally equipping them to manage most common and frequently occurring illnessesAs of the end of 2023, the ratio of general practitioners in China stands at approximately 3.99 per 10,000 people, falling short of the 5 per 10,000 goal set for 2030.
The introduction of DeepSeek presents a promising opportunity to elevate the standards of primary healthcareResearch indicates that the reinforcement learning framework of DeepSeek R1 facilitates autonomous learning and strategic adjustments in dynamic environmentsThis capability significantly enhances the model's adaptability to various clinical scenarios and patient needs.
Furthermore, a recent study conducted by a research team at Italy's University of Naples evaluated the pediatric clinical decision-support capabilities of ChatGPT and DeepSeek R1, emerging insights highlighting DeepSeek R1's evolving self-reflective capacity—enabling it to autonomously verify and optimize its logical processesThis feature can enhance performance for complex queries requiring multi-level analysis, such as determining subsequent management steps for children suspected of viral encephalitis.
In the future, industry models integrated with DeepSeek are likely to provide robust support for primary care physicians across various domainsFor instance, in diagnostics, general practitioners may sometimes struggle to make accurate assessments in the face of diverse and complex symptoms due to lack of experienceA medical model can integrate patient data such as symptoms, medical history, and test results to conduct comprehensive analyses, yielding precise diagnostic recommendations swiftly
Advertisements
Advertisements