Optic disc detection using fish school search algorithm based on FPGA

Abstract EN

Many people worldwide suffer from Diabetic Retinopathy (DP).

This health ailment affects their vision throughout the years, as they get older.

The fundus image is examined for detecting diabetic diseases that could affect the retina such as the DP.

Correctly detecting the optic disc is required to discover the disease.

Several methods have been proposed to improve the detection of the optic disc with respect to different performance metrics.

In this work, we investigate the performance, mainly the power consumption and the computational time, of the Fish School Search (FSS) technique.

We detect the optic disc by using contrast enhanced multi-step pre-processing technique to improve the color fundus image.

The pre-processing steps used in this work improve the quality of the colored image by filtering out the noises, smoothing the image, and masking out the regions where it is guaranteed that the optic disc is not located in.

The FSS algorithm is applied to find the brightest pixel in the pre-processed image, which is marked as the optic disc.

The algorithm is also implemented in the FPGA to benefit from the parallel processing power of the FPGA.

The algorithm is tested on DRIVE, STARE, and RiaRetDB1 databases and compared to other methods in the literature.

The accuracy of the FSS was 100%, 95.7%, and 99.92% when using DRIVE, STARE, and RiaRetDB1 databases, respectively.

Moreover, the running time of the FPGA implementation was found to be 1.605 ms with a total power dissipation of 121.818 mW.