Blood Vessel Segmentation in Retinal Fundus Images Using Hypercube NeuroEvolution of Augmenting Topologies (HyperNEAT)

Image recognition applications has been capturing interest of researchers for many years, as they found countless real-life applications. A significant role in the development of such systems has recently been played by evolutionary algorithms. Among those, HyperNEAT shows interesting results when dealing with potentially high-dimensional input space: the capability to encode and exploit spatial relationships of the problem domain makes the algorithm effective in image processing tasks. In this work, we aim at investigating the effectiveness of HyperNEAT on a particular image processing task: the automatic segmentation of blood vessels in retinal fundus digital images. Indeed, the proposed approach consists of one of the first applications of HyperNEAT to image processing tasks to date. We experimentally tested the method over the DRIVE and STARE datasets, and the proposed method showed promising results on the study case; interestingly, our approach highlights HyperNEAT capabilities of evolving towards small architectures, yet suitable for non-trivial biomedical image segmentation tasks.

[1]  Kenneth O. Stanley,et al.  Generating large-scale neural networks through discovering geometric regularities , 2007, GECCO '07.

[2]  Lawrence D. Jackel,et al.  Handwritten Digit Recognition with a Back-Propagation Network , 1989, NIPS.

[3]  Josh Harguess,et al.  Generative NeuroEvolution for Deep Learning , 2013, ArXiv.

[4]  Risto Miikkulainen,et al.  HyperNEAT-GGP: a hyperNEAT-based atari general game player , 2012, GECCO '12.

[5]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[6]  Yee Whye Teh,et al.  A Fast Learning Algorithm for Deep Belief Nets , 2006, Neural Computation.

[7]  Victor S. Lempitsky,et al.  N4-Fields: Neural Network Nearest Neighbor Fields for Image Transforms , 2014, ArXiv.

[8]  Shaun M. Lusk.  Evolving Neural Networks with HyperNEAT and Online Training , 2014 .

[9]  Josh Harguess,et al.  Feature Learning HyperNEAT: Evolving Neural Networks to Extract Features for Classification of Maritime Satellite Imagery , 2015, IPCAT.

[10]  Jan Cornelis,et al.  Analysis of a feature-deselective neuroevolution classifier (FD-NEAT) in a computer-aided lung nodule detection system for CT images , 2012, GECCO '12.

[11]  José Manuel Bravo,et al.  A New Supervised Method for Blood Vessel Segmentation in Retinal Images by Using Gray-Level and Moment Invariants-Based Features , 2011, IEEE Transactions on Medical Imaging.

[12]  Kenneth O. Stanley,et al.  A Hypercube-Based Encoding for Evolving Large-Scale Neural Networks , 2009, Artificial Life.