Hybrid human-artificial intelligence system for early detection and classification of AMD from fundus image

Imen Fourati Kallel,Sonda Kammoun

DOI: https://doi.org/10.1007/s11760-024-03115-2

2024-04-12

Abstract:In recent years, the use of machine learning and artificial intelligence for the diagnosis and monitor of various ocular disorders has been increasing remarkably. It has emerged as an indispensable diagnostic means of supporting practitioners to identify, detect and evaluate various diseases. This paper presents a computer-aided screening method, based on hybrid human-artificial intelligence, which detects and diagnoses the severity of age-related macular degeneration (AMD) from fundus image by using human–computer interaction. The specialist intervenes in the feature extraction phase, the preparation of training data, and the making of the decision result of the automatic algorithm through a synchronous communication with the system. The proposed automatic classification process is useful for the distinction of healthy from affected cases. Its accuracy reaches 98.4% by using texture features. Simultaneously, it helps to distinguish atrophic from neo-vascular AMD cases with classification accuracy of 96.34% by means of a combination of texture and colours features, surpassing a number of the state-of-the-art methods. The findings prove that the features extracted under the supervision of the specialists as a well-versed staff are a suitable incentive for the fundus image description.

engineering, electrical & electronic,imaging science & photographic technology