Multi-feature fusion-based strabismus detection for children

Strabismus is a common ophthalmologic disease that affects approximately 1.19% to 5.0% of children; however if the disease is detected early it can be treated effectively. Generally, the automatic detection of strabismus is usually performed only by a single feature, which is, with image deep features or ratio features. However, the accuracy of a strabismus diagnosis with a single feature is unreliable. This study aims to develop an intelligent strabismus detection model driven by corneal light reflection photographs to automatically detect children's strabismus. The proposed multi-feature fusion model (MFFM) improves the detection performance by fusing the deep features and ratio features extracted from the corneal light reflection photographs to identify strabismus. The experimental results demonstrate that our proposed multi-feature model outperforms all of the single feature models in strabismus detection. The experiments show that the proposed method achieves an accuracy of 97.17%, sensitivity of 96.06%, specificity of 97.79%, and AUC of 0.969 in strabismus detection. Our evidence shows that it greatly improves the performance of strabismus detection.