Convxtmob-Net: A Quantum -Secure Stacked Ensemble Model for Diabetic Retinopathy Severity Prediction and Classification

Abstract

Diabetic Retinopathy (DR) is one of the leading causes of vision loss worldwide. With proper and timely diagnosis, vision loss can be prevented. Especially when grading is automated to improve timely decision-making by physicians. In this work, a new quantum-secure stacked ensemble approach to the DR severity classification task is proposed. To tackle the data leakage issue in clinical grading, the ConvXtMob-Net accepts out-of-the-ordinary predictions and collaboratively learns from multiple models with various topologies. Provides a Module-Lattice-Based Key Encapsulation Mechanism (ML-KEM) for secure data transmission and simulated Quantum Key Distribution (QKD) for mathematically protecting proprietary model weights to tackle new security challenges in medical data transmission and edge development. ConvXtMob-Net implemented the solution and carried out a large number of experiments on the APTOS 2019 dataset, and the DDRs were 97.9% and 98.2%. Consistent Quadratic weighted kappa values of 0.97 testify to remarkable agreement in correspondence to the expert’s grading. The results indicate that the combination of stacked ensemble and post-quantum cryptography is a very precise, robust, and secure approach for automated clinical diagnostic systems.