Modified cross-over techniques in Genetic Algorithm for performance enhancement of retinal image classification system

Abnormality identification in retinal images has become one of the significant applications in the field of medical image analysis. Most of the detection techniques are computer aided which involves several computing techniques. One of the significant computing techniques is optimization which is extremely important for any abnormality detection system. Several optimization procedures are available among which Genetic Algorithm (GA) holds the significant position. But, conventional GA suffers from the drawback of randomness which degrades the performance of the overall image classification system. In this work, two Modified GA approaches are proposed to improve the performance of the system. These techniques are experimented on realtime retinal images collected from the scan centre. Feature extraction is initially performed followed by the implementation of the proposed optimization techniques for selecting the best features. Finally, image classification is performed with classification accuracy being the performance criteria for the proposed approaches. Experimental results show promising results for the proposed approaches.