减轻医生的压力-中兴通讯；中国电信

Relieving thePressure on Physicians Artificial intelligence streamlines and enhances the production ofcomprehensive medical reports Executive Summary •Demand is rising for comprehensive health checks, but delivering them can be both expensive andlabour-intensive for healthcare systems.•The senior physicians required to produce a comprehensive medical report have heavy workloadsand are in short supply•For individuals, it can be difficult to obtain detailed and comprehensible medical explanations andsuggestions.•To address these challenges, ZTE, Beijing Xingyun Digital Technology and Shanghai ShanzhenInformation Technology have integrated multidisciplinary knowledge and medical datasets to builda large AI model to automatically generate comprehensive medical reports.•Deployed at Shanghai Tenth People’s Hospital, the solution has increased daily report output tenfold.•The project partners are preparing to deploy the Intelligent Physical Examination Solution at scalefor medical service providers nationwide in China during 2026. medical examination solution. In April 2025, thepartners began engaging with Shanghai TenthPeople’s Hospital, and jointly conducted requirementanalysis and solution design. Thechallenge–risingdemandfor comprehensivemedicalchecks The hospital provided medical quality controldocuments and anonymised medical data. XingyunData Technology integrated this data withmultidisciplinary knowledge to build a largelanguage model that can support comprehensivemedical examinations. Shanghai Shanzhendeveloped the front-end application system forintelligent general inspection, while ZTE Corporationwas responsible for the AI server infrastructureand other facilities. For both individuals and healthcare providers,prevention is always better than cure, driving strongdemand for general health checks. However, to becomprehensive, such checks need to capture a lotof information, making them both an expensiveand labour-intensive bottleneck within healthcaresystems. As a typical physical examination department orinstitution in China usually has one or two chiefphysicians, at times of peak demand, thesephysicians often work overtime and stay up late,facing high work pressure and low efficiency. Drawing on sub-examination information, Xingyun’slarge language model can generate a comprehensivemedical report in a matter of minutes. The reportis then reviewed by physicians. For imagingexaminations (chest CT/DR, abdominalultrasound, etc.), laboratory examinations (bloodbiochemistry, etc.), and functional examinations(electrocardiogram, etc.), the large model generatesauxiliary diagnostic suggestions for doctors to referto when forming physical examination summaries.ZTE says the large model can also support otherprocesses such as: To produce a comprehensive medical report, a seniorphysician needs to review the sub-examinationresults of each medical department, conduct a fullanalysis based on the user’s health status, and finallyprovide summaries and recommendations. If eachreport takes 10-15 minutes to produce, a chiefphysician can output 30-50 reports per day. The reports typically list abnormal indicatorswithout analysing the causal relationships betweenthem or providing risk level classification, meaningthey can lack comprehensible medical explanationsand suggestions. As a result, they can leave patientsconfused or uninformed. —Quality control:Automatically verifying typos,unit errors, description completeness, genderlogic and conclusion consistency in physicalexam reports to ensure basic quality. —Abnormality handling:The large modelgenerates auxiliary diagnostic suggestions,automatically extracts abnormal info fromdepartment summaries and exam results, andproduces accurate standardised abnormalentries. Thesolution–training AIonextensive medical data —Suggestion generation:The system generatesprecise medical guidance and scientific reviewplans based on multi-dimensional data, suchas dynamic health risk ratings and thecorrelation analysis of abnormal indicators. To address these challenges, ZTE, Beijing XingyunDigital Technology and Shanghai ShanzhenInformation Technology have teamed up to harnessartificial intelligence (AI) to develop a comprehensive —Health risk assessment:The model cansynthesise current severity, indicator correlationand longitudinal trends to generate high/medium/low risk dynamic ratings, and specifymain risk sources. Provided by China Telecom, theprivate 5G network encompassesin-hospital base stations anddistributed antenna systems,which relay medical datadirectly to the hospital’s coreswitch without passing throughpublic networks. The data isprocessed in an AI studio runningon AI servers located on-premise –the hospital’s private cloud.This edge architecture is designedto keep health data secure andprivate. —In-depth analysis and insight:Based on medicalknowledge graphs, the system analyses theinternal clinical associations betweenabnormal indicators (e.g., the correlationbe