Page 144 - AI for Good-Innovate for Impact Final Report 2024

AI for Good-Innovate for Impact

and constructing intelligent service processes for the pre diagnosis process; Targeting medical
research, building clinical research data governance capabilities and clinical research indicator
systems, providing data retrieval services, data processing services, intelligent indicator
development services, and research output assistance to support clinical research scenarios
for doctors; Develop the ability to interpret health reports for health examination scenarios.
Build an intelligent service loop centered around clinical testing, including pre examination, in
examination, and post examination, to provide higher quality and flexible AI medical services
for intelligent healthcare.

The current plan utilizes the understanding and generation capabilities of large models, which
can help doctors and patients quickly understand the condition, assist in health assessment,
department recommendations, and so on. Compared to the past, using information databases
for retrieval or deep learning model solutions is faster, more convenient, and more user-friendly.
However, the current large model still has the problem of "hallucinations", which can only be
auxiliary and cannot be used for decision-making.

GPU: Ascend910B

Use case status: The use case is part of a larger research project

UN Goals:
• SDG 3: Good Health and Well-being

SDG3 - Good Health and Well-being: Yuanjing Large Model (Medical) is an industry model
specifically trained for the medical industry. It can provide patients with functions such as
medical selection, pre diagnosis information collection, and pre consultation before seeking
medical treatment, provide department recommendation services during medical treatment,
and provide health report interpretation services after medical treatment to help patients seek
medical treatment more conveniently and flexibly. At the same time, it can provide clinical
research data retrieval and processing services for doctors, as well as auxiliary services such
as research output and paper analysis for medical research, helping doctors better engage in
medical and scientific research work.

32�2�2 Future work

Create new variations/extensions to the same use case

Elaborate proposal:

• Completed in the first half of 2024: recommended departments for medical treatment,
recommended hospitals, natural language retrieval for medical research assistants,
analysis of test results, interpretation of multimodal reports, pre - and post medical
inquiries from patients. And coordinate with local hospitals for initial implementation.
• Completed in the second half of 2024: pre consultation health assessment, medical
research assistant scientific research data analysis and statistics, paper analysis, assisted
paper writing, automatic output of general inspection reports, drug inquiry, drug
knowledge Q&A, medication guidelines, personalized follow-up, health assessment and
other services. Further promote the application scope

In the future, we will further iterate our product capabilities based on market and medical
patient needs, and promote them to more hospitals.

128