Open Access Open Access  Restricted Access Subscription Access

VisionCare AI: An Explainable AI-Based Retinal OCT Analysis and Clinical Decision-Support System

Mr. Mohamed Arafath A, Mrs. Deepthi Nair P, Dr. Subha Hency Jose, Dr. Anantha Christu Raj P

Abstract


Optical Coherence Tomography (OCT) is the clinical gold standard for non-invasive, micrometer-resolution cross-sectional visualization of retinal architecture. Rapid and accurate identification of sight-threatening macular diseases—specifically Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), and Drusen—is vital for initiating prompt therapeutic interventions, such as anti-vascular endothelial growth factor (anti-VEGF) injections, to arrest irreversible visual loss. However, manual interpretation of large-scale volumetric B-scans places an unsustainable cognitive burden on vitreoretinal specialists, suffers from observer subjectivity, and is severely restricted by the shortage of ophthalmologists in low-resource settings. In this work, we present VisionCare, an explainable, end-to-end artificial intelligence workstation designed for automated macular condition classification, pathological lesion segmentation, and quantitative biomarker extraction from spectral-domain OCT (SD-OCT). The diagnostic core integrates a Dual-Stream Neural Consensus Engine combining a compound-scaled EfficientNet-B0 network with an inverted-residual MobileNet-V3 backbone. This consensus mechanism is governed by an information-theoretic Shannon Entropy Epistemic Uncertainty Filter that flags ambiguous or borderline scans and autonomously routes them to human experts.

Simultaneously, an Attention U-Net segmentation network isolates pathological fluid accumulations—including intraretinal fluid (IRF), subretinal fluid (SRF), and pigment epithelial detachments (PED)—to derive localized lesion masks and Central Subfield Thickness (CST) indices aligned with the 9-zone Early Treatment Diabetic Retinopathy Study (ETDRS) standard. Model explainability is provided through Gradient-weighted Class Activation Mapping (Grad-CAM), ensuring that deep representations directly correspond to visible pathological changes. Evaluated on the benchmark UCSD retinal OCT dataset (  scans), the proposed consensus framework achieves an overall classification accuracy of , sensitivity of , specificity of , and an Area Under the Receiver Operating Characteristic curve (AUC-ROC) of . The segmentation network achieves a Dice Similarity Coefficient (DSC) of with an end-to-end pipeline latency of . Integrated into an operational clinical workstation featuring a PACS-grade curtain slider, digital micrometer calipers, bilingual clinical briefing synthesis (English and Tamil), and retrieval-augmented conversational assistance, VisionCare delivers a reliable, transparent, and scalable tool for computer-aided ophthalmological diagnosis.


Full Text:

PDF

References


Kermany, D. S., et al., "Identifying Medical Diagnoses and Treatable Diseases by Image-Based Deep Learning," Cell, vol. 172, no. 5, pp. 1122–1131, 2018.

Tan, M., and Le, Q. V., "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks," in International Conference on Machine Learning (ICML), 2019, pp. 6105–6114.

Howard, A., et al., "Searching for MobileNetV3," in IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 1314–1324.

Oktay, O., et al., "Attention U-Net: Learning Where to Look for the Pancreas," in Medical Image Computing and Computer Assisted Intervention (MICCAI), 2018, pp. 1–10.

Selvaraju, R. R., et al., "Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization," in IEEE International Conference on Computer Vision (ICCV), 2017, pp. 618–626.

Schlegl, T., et al., "Fully Automated Detection and Quantification of Macular Fluid in OCT Using Deep Learning," Ophthalmology, vol. 125, no. 4, pp. 549–558, 2018.

Rasti, R., et al., "Macular OCT Classification Using a Multi-Scale Convolutional Neural Network with Shared Weights," IEEE Transactions on Medical Imaging, vol. 37, no. 9, pp. 1964–1974, 2018.

Sotoudeh-Paima, A., et al., "Multi-Scale Convolutional Neural Network for Retinal Optical Coherence Tomography Image Classification," Computers in Biology and Medicine, vol. 144, p. 105373, 2022.

Shannon, C. E., "A Mathematical Theory of Communication," The Bell System Technical Journal, vol. 27, no. 3, pp. 379–423, 1948.

Early Treatment Diabetic Retinopathy Study Research Group, "Grading Diabetic Retinopathy from Stereoscopic Color Fundus Photographs: An Extension of the Modified Airlie House Classification (ETDRS Report Number 10)," Ophthalmology, vol. 98, no. 5, pp. 786–806, 1991.


Refbacks

  • There are currently no refbacks.