Development and Deployment of DeepSeek-Based Applications in Healthcare: A Chinese Perspective

Area of Science:

• Healthcare Technology
• Artificial Intelligence in Medicine
• Digital Health Transformation

Background:

• Generative AI accelerates digital transformation in healthcare, offering decision support, automated diagnosis, and resource optimization.
• DeepSeek-R1, a cost-effective large language model, is gaining traction in Chinese hospitals.
• Hospitals face challenges in selecting generative AI deployment architectures and balancing computational demands with costs.

Purpose of the Study:

• To survey AI deployment in Chinese hospitals, focusing on DeepSeek's applications within national policies.
• To analyze deployment strategies, model versions, and platform choices based on hospital needs and resources.

Main Methods:

• Survey of AI deployment in top-tier, regional, and township hospitals in China via official WeChat platforms.
• Examination of deployment strategies, model versions (e.g., 671B, 70B, 32B), and platform choices.
• Consideration of hospital needs, data resources, and technological-economic factors.

Main Results:

• DeepSeek demonstrates impact on diagnostic accuracy, personalized treatment, medical documentation, and resource management.
• Specific model versions (671B, 70B, 32B) were adopted by surveyed hospitals.
• Hospitals generally preferred local deployment, with varied needs and applications observed.

Conclusions:

• Optimal AI model selection requires balancing computational power and cost; larger models offer accuracy but higher costs.
• Distilled models are suitable for smaller hospitals with limited resources.
• Future AI development should align deployment strategies with hospital size and address data quality to reduce healthcare disparities.