The Role Of Deep Learning Algorithms In Cataract Detection And Surgical Planning Just Through Smartphone Eye Images

To explore the potential of smartphone-based imaging and artificial intelligence (AI) in the detection and grading of cataracts, with an emphasis on optimizing the timing of surgical intervention. The study aims to highlight how noninvasive, accessible diagnostic methods can enhance early detection and improve patient outcomes.

This study was conducted in Vision health research clinic utilizing smartphone imaging combined with AI deep-learning models for cataract analysis and grading.

A dataset of cataract images was collected using high-resolution smartphone cameras (S23 Samsung Mobile phone) under standardized lighting conditions. The dataset included both normal (1021 images) and cataract eyes (1400 images), categorized by severity . Deep learning algorithms were trained with 80% of datas, to detect lens opacity, assess its progression in monitoring grades, and provide a grading score based on severity. The AI model’s accuracy was validated against clinical grading performed by two independent ophthalmologists and a labeling expert.

The AI system demonstrated a detection accuracy of 93.2%, with a sensitivity of 91.2% and a specificity of 94.5% in distinguishing cataractous from normal eyes. The area under the receiver operating characteristic (ROC) curve (AUC) was 0.96, indicating high diagnostic reliability. The model successfully categorized cataracts into early, moderate, and advanced stages. The confusion matrix analysis confirmed strong agreement with human expert grading. Furthermore, the AI-assisted grading system facilitated improved preoperative decision-making by providing a consistent and objective assessment of cataract severity, enabling optimized timing for surgical intervention and reducing postoperative complications.

Smartphone-based imaging, combined with AI-powered analysis, presents a cost-effective and scalable solution for early cataract detection and grading anywhere, anytime. By integrating this technology into routine ophthalmic assessments, clinicians can enhance preoperative decision-making and optimize surgical timing, ultimately improving sustainable patient care. Future advancements should focus on refining AI algorithms and expanding accessibility to underserved regions.