Detection of Strabismus Using Convolutional Neural Network-Based Classification Models

Strabismus, a prevalent eye misalignment disorder, can lead to vision impairment if not promptly identified and treated. Existing automatic detection approaches often rely on single-feature methodologies, such as deep features or ratio features, which may suffer from reliability issues. This paper introduces a pioneering multi-feature fusion model (MFFM) aimed at augmenting the accuracy of strabismus detection that integrates deep features and ratio features extracted from corneal light reflection photographs using convolutional neural network (CNN) based on classification models. Initially, deep features are extracted employing pre-trained CNN models renowned for its efficacy in diverse image recognition and classification of strabismus images. Through the fusion of these features, MFFM offers a robust framework to bolster the accuracy of strabismus detection. This research presents a promising avenue for advancing early diagnosis and treatment of strabismus, thereby potentially mitigating the risks of vision impairment.