Supervised vessel segmentation with minimal features

• Main Authors: Che Azemin, Mohd Zulfaezal, Mohd Tamrin, Mohd Izzuddin
• Format: Conference or Workshop Item
• Language: English
• Published: 2014
• Subjects: RE Ophthalmology; TK7800 Electronics. Computer engineering. Computer hardware. Photoelectronic devices; TK7885 Computer engineering
• Online Access: http://irep.iium.edu.my/42183/4/su.pdf; http://irep.iium.edu.my/42183/; http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=7036744

Current state-of-the art supervised vessel segmentation methods require large number of feature vectors to construct a good model. In this paper, we propose a framework to optimally search for optimal features as inputs to Artificial Neural Network (ANN) trained by Scaled Conjugate Gradient (SCG). SCG is known to speed-up the learning stage in a supervised learning especially when error reduction is critical. The proposed framework is able to reduce features from 16 to 4 dimensions and the overall performance is only decreased by 1% in average