A VR-based Automated Strabismus Diagnosis System with Progressive Semi-Supervised Learning.

Area of Science:

• Ophthalmology
• Computer Science
• Medical Imaging

Background:

• Strabismus is a common eye disorder impacting visual development and psychological well-being.
• Conventional diagnosis using the prism cover test (PCT) is subjective and lacks standardization.
• Automated diagnosis faces challenges in VR simulation, image degradation, and limited annotated data.

Purpose of the Study:

• To develop a novel virtual reality (VR)-based automated strabismus diagnosis system.
• To address challenges in realistic VR simulation, image degradation, and precise ocular deviation quantification.
• To introduce a new clinical dataset (TongRenD) and a robust semi-supervised deep learning model (ProgNet).

Main Methods:

• Developed a VR framework with five standardized clinical examination scenarios.
• Introduced ProgNet, an uncertainty-guided progressive semi-supervised segmentation network with a Prototype-based Feature Representation Module (PFRM).
• Implemented a 3D deviation estimation algorithm for strabismus classification and angular measurement.

Main Results:

• ProgNet demonstrated superior segmentation accuracy compared to state-of-the-art methods on TongRenD and TEyeD datasets.
• The system achieved high consistency with expert assessments during clinical validation.
• The developed system provides a standardized, non-invasive, and reliable solution for strabismus diagnosis.

Conclusions:

• The proposed VR-based automated system offers a significant advancement in strabismus diagnosis.
• The ProgNet model effectively handles image degradation and limited annotations.
• This technology has the potential to improve early diagnosis and management of strabismus.