Predicting Intraocular Collamer Lens Vault in Myopic Patients With Shallow Anterior Chamber Using FedEYE Platform and...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Shallow anterior chamber depth (ACD) presents challenges in implantable collamer lens (ICL) surgery.
• Accurate prediction of postoperative vault is crucial for optimizing surgical outcomes and patient safety.

Purpose of the Study:

• To evaluate the effectiveness of code-free artificial intelligence (AI) platforms, FedEYE and ChatGPT, in predicting vault in patients with shallow ACD undergoing ICL surgery.
• To compare the performance of different AI models and data combinations for vault prediction.

Main Methods:

• A retrospective study of 160 eyes with shallow ACD (<2.8 mm) was conducted.
• Two code-free AI models were developed: an integrated neural network (INN) using FedEYE with optical coherence tomography (OCT) images and examination data, and ChatGPT utilizing memory prompting.
• The models were trained and validated using multimodal preoperative data.

Main Results:

• The INN model, combining OCT images and examination data, achieved high accuracy (0.875), sensitivity (0.857), specificity (0.889), and AUC (0.873).
• Anterior chamber angle and spherical equivalent positively correlated with vault; age and white-to-white diameter showed negative correlation.
• ChatGPT, after memory prompting, demonstrated good performance with accuracy of 0.813 and AUC of 0.865.

Conclusions:

• Code-free AI tools like FedEYE and ChatGPT can be developed and deployed straightforwardly.
• These AI models show encouraging performance in predicting vault for shallow ACD patients, aiding surgical planning.
• Utilizing AI for vault prediction holds significant clinical potential for enhancing safety and efficacy in ICL surgeries for this patient group.