Automatic selection of parameters for vessel/neurite segmentation algorithms

Abstract

An automated method is presented for selecting optimal parameter settings for vessel/neurite segmentation algorithms using the minimum description length principle and a recursive random search algorithm. It trades off a probabilistic measure of image-content coverage against its conciseness. It enables nonexpert users to select parameter settings objectively, without knowledge of underlying algorithms, broadening the applicability of the segmentation algorithm, and delivering higher morphometric accuracy. It enables adaptation of parameters across batches of images. It simplifies the user interface to just one optional parameter and reduces the cost of technical support. Finally, the method is modular, extensible, and amenable to parallel computation. The method is applied to 223 images of human retinas and cultured neurons, from four different sources, using a single segmentation algorithm with eight parameters. Improvements in segmentation quality compared to default settings using 1000 iterations ranged from 4.7%-21%. Paired t-tests showed that improvements are statistically significant (p < 0.0005). Most of the improvement occurred in the first 44 iterations. Improvements in description lengths and agreement with the ground truth were strongly correlated (p = 0.78).

A Cognitive Vision Approach to Image Segmentation

“…Almost all of the free derivative optimization techniques have been tested. Interesting frameworks can be found in (Bahnu et al, 1995;Peng & Bahnu, 1998;Mao et al, 2000;Cinque et al, 2002;Gelasca et al, 2003;Pignalberi et al, 2003;Abdul-Karim et al, 2005). In the worst case, results of optimized segmentations are equivalent to the ones obtained with default parameters.…”

A Statistical Model for Recreational Trails in Aerial Images

“…For example, biomedical researchers and clinicians need to trace neurons [2,3] and blood vessels [1]. They have used a variety of approaches, such as active contours [21], minimal paths [9], and many other methods.…”

A Learning Approach for Adaptive Image Segmentation

“…The proposed optimization procedure overcomes such limitations by decomposing the problem into three fundamental and independent components: a segmentation algorithm with its free-parameters to tune, a segmentation evaluation metric and a global optimization algorithm (see Figure 15). To our knowledge, this scheme has already been applied for adaptive segmentation problems by Banu et al (Bahnu et al, 1995) and by Abdul-Karim et al (Abdul-Karim et al, 2005). Bahnu et al used a genetic algorithm to minimize a multiobjective evaluation metric based on a weighted mix of global, local and symbolic information.…”