Development of a Novel Slit-Lamp Replication and ResNet Platform for Early Non-Invasive Detection of Glaucoma and Kayser-Fleischer Rings in Neurological Wilson’s Disease and Liver Failure.
Wilson’s disease is an autosomal recessive disorder causing toxic copper accumulation. A key diagnostic feature is Kayser-Fleischer (KF) rings, which are often missed in early stages, especially in darker irises. Traditional slit-lamp exams require expensive equipment ($5,000–$10,000) and specialists, limiting global access. EyeNova bridges this gap with an AI-driven, accessible diagnostic tool.
Inspired by the mantis shrimp—which detects 16 distinct photoreceptor channels from UV to IR—EyeNova expands perception beyond human trichromatic vision. The pipeline applies intensity scaling, channel-wise transformations, and integrates infrared (IR) imaging to amplify subtle chromatic differences that are typically invisible.
Modeled in Fusion 360 and 3D-printed for rapid prototyping, the system operates on a Raspberry Pi with an IR camera. Iteration V2 features an organic, eye-shaped eyecup for optimal focus, expanded housing for status LEDs, and improved airflow. Total hardware cost remains radically accessible at under $50.
Two modified ResNet-50 Convolutional Neural Networks were trained using TensorFlow. By augmenting conventional RGB inputs with IR data and multispectral transformations, the deeper convolutional layers can extract high-dimensional features simulating true multispectral perception.
Rigorous testing proved the system's clinical viability. The Glaucoma model achieved 89.40% accuracy (92.6% sensitivity), while the KF Ring model reached an exceptional 93.43% accuracy, significantly outperforming average human diagnostic rates (60–70%). EyeNova offers a highly scalable, point-of-care solution that democratizes specialized ophthalmic screening.