DenSFFNet: dense spiking forward fractional network for cardiovascular risk prediction using retinal fundus images in spark framework

Cardiovascular risk prediction identifies individuals at high risk before symptoms arise. To address challenges such as integrating diverse data, ensuring quality, and managing patient variability, the Dense Spiking Forward Fractional Network (DenSFFNet) model is introduced within the Spark framework. The process begins with image acquisition and partitioning using Deep Embedded Clustering (DEC), followed by preprocessing tasks like Greyscale Conversion, Optic Disc (OD) segmentation with Channel Prior Convolutional Attention (CPCA), and blood vessel segmentation using Frangi-Net across slave nodes. Extracted features, including Learned Invariant Feature Transformation (LIFT) and statistical metrics, are aggregated by the master node, which utilises the DenSFFNet model a combination of DenseNet and Deep Spiking Neural Network (DSNN). The DenSFFNet method attained accuracy, sensitivity, specificity, and Matthews correlation coefficient (MCC) is 91.119%, 90.366%, 89.922%, and 92.643% for dataset 1. For the RFMiD 2.0 dataset, the proposed method attained 90.881% accuracy, 90.286% sensitivity, 89.660% specificity, and 91.469% MCC.