| Abstract [eng] |
Artificial Intelligence Assisted Diagnostic Procedures in Ophthalmology Artificial intelligence has the potential to transform diagnostic procedures in medicine. The use of this tool is applicable in ophthalmology since the diagnosis of eye diseases is based on image or visual tests. This review aims to assess the accuracy of artificial intelligence-based diagnostic tools, provide an overview of the principles of artificial intelligence, and discuss the advantages and disadvantages of this diagnostic modality. A search of the PubMed database was carried out and 38 publications were selected for the systematic review and 27 publications for the literature review. The systematic review included 20 publications on diabetic retinopathy, 8 publications on glaucoma, 5 publications on age-related macular degeneration, and 5 publications on other eye diseases. Diagnostic accuracy was analyzed using the area under the curve (AUC), sensitivity and specificity. In some of the publications scientists also compared diagnostic accuracy between artificial intelligence-based devices and medical specialists. For diabetic retinopathy, the mean AUC was 0.940, with a sensitivity of 93.82 per cent and a specificity of 90.48 per cent. Glaucoma had a mean AUC of 0.933, a sensitivity of 86.90 per cent, and a specificity of 90.16 per cent. Age-related macular degeneration had a mean AUC of 0.960, a sensitivity of 90.02 per cent, and a specificity of 91.00 per cent. Thus, artificial intelligence is a rapidly advancing technology that has great potential to enter clinical practice in ophthalmology. The accuracy of the artificial intelligence-based diagnostic systems in this analysis was good enough to use these devices in clinical practice. Publications have also shown that the diagnostic accuracy of artificial intelligence-based devices has been equal to or better than that of medical specialists. |
