Skip to content

A real-world retrospective validation study of a large model for ophthalmic auxiliary diagnosis driven by multi-source clinical texts

A real-world retrospective validation study of a large model for ophthalmic auxiliary diagnosis driven by multi-source clinical texts

Status
Active, not recruiting
Phases
Unknown
Study type
Observational
Source
ChiCTR
Registry ID
ChiCTR2600127117
Enrollment
Unknown
Registered
2026-06-24
Start date
2026-07-01
Completion date
Unknown
Last updated
2026-06-29

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Ophthalmic diseases

Interventions

NA:NA

Sponsors

Shanghai General Hospital
Lead Sponsor

Eligibility

Sex/Gender
All

Inclusion criteria

Inclusion criteria: 1. The patients who were hospitalized in the Ophthalmology Department of Shanghai First People's Hospital during the study time window; The research time window is tentatively set from January 1, 2025 to December 31, 2025. Before submission, it can be adjusted according to the extractable range of the Information Department. 2. The presence of clear major ophthalmic diagnoses or major ophthalmic problems in the inpatient medical records can form a reference standard diagnosis. 3. The medical record should contain at least one resolvable source of ophthalmic phenotype information, including but not limited to: chief complaint/present illness history, specialized physical examination, examination reports, surgical records, and pathological reports. 4. Medical records can be de-identified, and necessary research information is still retained after desensitization.

Exclusion criteria

Exclusion criteria: 1. The core medical records are seriously lacking, making it impossible to construct a minimum case summary. 2. Those who only reflect inpatient management or perioperative status and cannot determine the diagnosis of underlying eye diseases. 3. Those who have been repeatedly extracted, whose patient identities cannot be duplicated, or who have been repeatedly extracted from the same hospitalization. 4. Records that have a clear risk of desensitization failure or are assessed by the research team as unsuitable for inclusion in the analysis. 5. Ultimately, the diagnostic label standardization aligner cannot be completed (it can enter the coverage analysis, but not the main diagnostic performance analysis).

Design outcomes

Primary

MeasureTime frame
Medical record information of inpatients;Model evaluation indicators;

Countries

China

Contacts

Public ContactTong Li

Shanghai General Hospital

019139lt@shgh.cn+86 21 63240090

Outcome results

None listed

Source: ChiCTR (via WHO ICTRP) · Data processed: Jul 3, 2026