Free Papers

Artificial intelligence in biometrical macular scan to detect macular disease in patients scheduled for cataract surgery

Free Paper Details

First Author: M.Pastore ITALY

Co Author(s):    R. Merli   L. Belfanti   G. Cirigliano   M. Franzon   D. Tognetto        

Abstract Details

Purpose:

To create a normative database based on pattern recognition as a reference model introduced in an optical biometer report to standardize the foveal pit analysis offered by the swept-source optical coherence tomography (SS-OCT).

Setting:

Eye Clinic, Department of Medical, Surgical Sciences and Health, University of Trieste, Trieste, Italy

Methods:

Prospective observational study of 2352 eyes of 1491 consecutive patients scheduled for cataract surgery. The preoperative examination included a biometrical examination with SS-OCT IOLMaster 700 and SD-OCT device analysis (Spectralis OCT, Heidelberg Engineering, Heidelberg, Germany) on the same day. An independent external operator exported anonymously all macular imaging. Four readers graded the SS-OCT scan as normal or pathological, using SD-OCT imaging for not clear scans.

Results:

SD-OCT examination revealed a macular pathology in 938 eyes (41.3%), with a normal macular scan in 1333 eyes (58.7%). 346 over 2352 biometrical scans were graded not evaluable for media opacities or artifacts during the examination and were removed from further investigation. All the 2006 images were pre-processed to obtain reproducibility of the results even for analysis performed with biometric foveal images of any size and format. Using 65% of dataset (1304 scans) as train and 35% (702 scans) as a test, a neural network was trained to identify normal and pathological foveal profile with an accuracy of 90%.

Conclusions:

Artificial intelligence represents an important contribution to the optimization of routine activities. The introduction of a normative database of the SS-OCT macular biometer scan provides useful information concerning the macular structure. The optical biometer SS-OCT imaging could be an effective, helpful modality to predict the risk of reduced visual recovery and to reduce the need to perform the conventional SD-OCT examination in all patients scheduled for cataract surgery.

Financial Disclosure:

-