Page 48 - AI for Good Innovate for Impact
P. 48

AI for Good Innovate for Impact



                      (continued)

                       Item          Details
                       Technology    RAG, high-quality medical datasets, AI agents
                       Keywords
                       Data Availabil- RJUA-QADatasets: Ant-Renji Urology Specialty QA Dataset [12]
                       ity

                       Metadata      structured and unstructured data
                       (Type of Data)

                       Model Train- The model is pretrained on high-quality, domain-specific medical data, includ-
                       ing      and ing both textual and multimodal sources such as visually rich clinical documents.
                       Fine-Tuning   The model is then finetuned using expert annotated QA pairs and multi-turn
                                     dialogues which adapted to various downstream medical application scenarios.

                       Testbeds or  Zhejiang Provincial Health Commission [13]
                       Pilot  Deploy-  Hangzhou Municipal Medical Insurance Bureau [14]
                       ments
                                     Shanghai Jiao Tong University School of Medicine, Renji Hospital [15]

                       Code reposi- Code repositories are available [16]  [17]
                       tories



                      2      Use Case Description


                      2�1 Description


                      Challenges
                      •    Patient experience compromised due to the gap between supply and demand for
                           medical resources: As living standards rise, the demand for healthcare services has
                           increased in both quantity and quality, but the growth of medical resources has not fully
                           kept pace with this change.
                      •    Complex and administrative workflow limits doctors' clinical focus: In traditional healthcare
                           sector, doctors have to deal with complex workflows, limiting the time they can dedicate
                           to diagnosis and treatment.
                      •    Uneven geographic  distribution of  medical  resources:  There  are  disparities  in  the
                           geographic distribution of healthcare resources, affecting the accessibility and
                           convenience of high-quality healthcare for residents in different regions.

                      These factors combined have impacted on the level of healthcare available to residents across
                      different areas.


                      Solutions

                      To address the above challenges, this case study proposes an intelligent application solution
                      based on Ant Group's large medical model, aiming optimize healthcare resources through
                      technology, promote universal access to smart healthcare services, and advance wellbeing
                      for all.

                      •    For Residents: This solution introduces an AI health assistant that covers the entire
                           medical journey, allowing everyone to have a smart family doctor. For patients with
                           explicit healthcare needs, AI health assistant can provide comprehensive guidance and



                  12
   43   44   45   46   47   48   49   50   51   52   53