AI reading of IOLMaster 700 reports to recommend toric lenses and refractive targets
Head-to-Head Evaluation of ChatGPT 4o, GPT-5, and DeepSeek for Structured Extraction, Toric IOL Recommendation, and Refractive Prediction
This project tests whether large language models can read de-identified IOLMaster 700 report images to pull out eye measurements, recommend toric intraocular lenses, and predict postoperative refractions for people with cataract.
Quick facts
| Study type | Observational |
|---|---|
| Enrollment | 100 (estimated) |
| Ages | 18 Years and up |
| Sex | All |
| Sponsor | Eye & ENT Hospital of Fudan University Academic / other |
| Locations | 1 site (Shanghai, Shanghai Municipality) |
| Trial ID | NCT07183891 on ClinicalTrials.gov |
What this trial studies
This was a single-center, retrospective observational analysis using de-identified IOLMaster 700 report images from 54 participants (162 eyes). Three large language models (ChatGPT 4o, GPT-5, and DeepSeek) were supplied native, unprocessed raster images and prompted to (i) extract structured biometric parameters, (ii) give a binary toric candidacy decision and an institutional T-code, and (iii) provide refractive recommendations (sphere, cylinder, axis). Each model produced three independent outputs per exam; primary outcomes were parameter-level agreement and refractive error metrics, and secondary outcomes included decision-support performance for toric IOL selection and T-code agreement. No clinical interventions were performed, and the work proceeded under IRB oversight with de-identification and a waiver of consent.
Who should consider this trial
Good fit: Ideal candidates are people with cataract who have complete IOLMaster 700 biometric reports and who had uncomplicated surgery with good early postoperative visual acuity and rotational stability per the protocol.
Not a fit: Patients with incomplete or unreadable biometric reports, prior ocular surgery or trauma, intraoperative or significant postoperative complications, or atypical anatomy are unlikely to benefit from the model outputs.
Why it matters
Potential benefit: If successful, this approach could speed and standardize biometric extraction and provide automated decision support for toric IOL selection and refractive planning, reducing manual review time.
How similar studies have performed: Previous AI and machine-learning efforts have shown promise for extracting biometry and improving IOL calculations, but using large language models directly on raw report images for toric decision support is relatively novel and less validated.
Eligibility criteria
Show full inclusion / exclusion criteria
Inclusion Criteria: -postoperative corrected distance visual acuity (CDVA) of 0.10 logMAR or better -an absolute IOL rotational stability of less than 10∘ at the 1-month follow-up examination Exclusion Criteria: * incomplete biometric data on the examination report; * a history of previous ocular surgery or ocular trauma * the occurrence of intraoperative complications, such as an anterior capsular tear or posterior capsular rupture * the development of significant postoperative complications, including but not limited to severe intraocular infection or inadequate pupillary dilation.
Where this trial is running
Shanghai, Shanghai Municipality
- Eye and ENT hospital of Fudan University — Shanghai, Shanghai Municipality, China (Recruiting)
Study contacts
- Study coordinator: Xuanqiao Lin
- Email: 1532483480@qq.com
- Phone: +8615088920668
How to participate
- Review the eligibility criteria above with your treating physician.
- Visit the official trial page on ClinicalTrials.gov for the most current contact information and recruitment status.
- Contact the listed study coordinator or principal investigator to request pre-screening. Pre-screening is free and never obligates you to enroll.