•  
  •  
 

Abstract

Background: Artificial intelligence (AI) has increasingly been integrated into various aspects of healthcare, revolutionizing diagnostics, patient management, and medical decision-making. AI-powered platforms, particularly large language models (LLMs) such as ChatGPT, have demonstrated potential in supporting clinical expertise by generating relevant, context-aware responses to medical inquiries. In ophthalmology, AI has been widely utilized for diagnosing and monitoring common conditions like glaucoma, diabetic retinopathy, and macular degeneration. However, its application in strabismus management, particularly in surgical planning, remains underexplored.

 Objective: to compare the accuracy of responses about surgical plans for strabismus by ChatGPT-3.5 and ChatGPT-4 with the traditional surgical strabismus table of Parks for exodeviation to amount of Bilateral Lateral Rectus Recession (BLR) has been already done by a strabismus consultant in cases of intermittent exotropia (IXT).

Patients and methods: We started by asking ophthalmologists to complete a survey to find out how well-versed they are in using ChatGPT or other AI tools for clinical decision-making and surgical planning then all of the patient's ocular and preoperative strabismus examination data of 30 strabismus surgeries (BLR) carried out by the same strabismus consultant in cases of IXT which is characterized by free ocular motility without any muscle restriction or overaction prior to surgery and they developed orthophoria after surgery were submitted in the ChatGPT-3.5 and ChatGPT-4 as (1st  response) and were submitted by same input in both versions again after 2 weeks as (2nd response)  then the accuracy of both responses ( 1st , 2nd   responses ) regarding surgical plans (amount of BLR ) for these cases provided by both versions of ChatGPT were compared to the numbers in the classical surgical strabismus table of Parks for exodeviation and the amount already done for these cases by the same strabismus consultant..

Results:The survey, conducted among 42 ophthalmic professionals, revealed a moderate adoption of AI in clinical practice with Chat GPT was the preferred search engine (62%) by more than half  , perceptions of AI accuracy are mixed, with 43% rating it as somewhat accurate, while 17% found it somewhat inaccurate. Regarding future AI use in surgical planning, more than half (55%) expressed willingness to adopt it if proven reliable, while 26% remained skeptical , then the total of 30 cases of intermittent exotropia were analyzed, and the response accuracy of AI-based recommendations was assessed showed that a statistically significant difference (p

Conclusion: According to the survey of 42 ophthalmic professionals indicates a moderate use of AI in clinical practice; more than half of them chose ChatGPT as their preferred search engine by (62%). And according to our study results indicate that while AI-generated surgical plans for strabismus show some degree of accuracy, there are still differences, especially with ChatGPT-4's recommendations, and that responses differed when it has been asked about the same cases 2 weeks later with the same inputs, indicating inconsistent AI-generated recommendations in both versions of Chat GPT. Compared to ChatGPT-4, ChatGPT-3.5 however showed a minimal stronger alignment with expert recommendations ,numerous results still displayed distant numbers ; more than  0.5 mm sometimes . This suggests that even with AI's progress, more work is required to increase the accuracy of AI-assisted surgical decision-making in ophthalmology , and Chat GPT is still unable to make a decision regarding the surgical dosage of BLR in cases of intermittent exotropia and we do not recommend depending primarily on ChatGPT's recommendations when determining surgical dosages for strabismus surgeries

Article Type

Original Study

Subject Area

Ophthalmology

IRB Number

MIOR0035

Creative Commons License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial-Share Alike 4.0 International License.

Additional Files
Questionnaire figure.docx (135 kB)
Evaluating the Accuracy of ChatGPT in Decision of Surgical Dose i.pdf (984 kB)
After Review and Revise Article Proof [JMISR_1543]
Download

Share

COinS