Discriminative Parameter Estimation for Random Walks Segmentation

Abstract

The Random Walks (RW) algorithm is one of the most efficient and easy-to-use probabilistic segmentation methods. By combining contrast terms with prior terms, it provides accurate segmentations of medical images in a fully automated manner. However, one of the main drawbacks of using the RW algorithm is that its parameters have to be hand-tuned. we propose a novel discriminative learning framework that estimates the parameters using a training dataset. The main challenge we face is that the training samples are not fully supervised. Specifically, they provide a hard segmentation of the images, instead of a probabilistic segmentation. We overcome this challenge by treating the optimal probabilistic segmentation that is compatible with the given hard segmentation as a latent variable. This allows us to employ the latent support vector machine formulation for parameter estimation. We show that our approach significantly outperforms the baseline methods on a challenging dataset consisting of real clinical 3D MRI volumes of skeletal muscles.

Graph-Based Slice-to-Volume Deformable Registration

“…The integration of geometric information (landmark correspondences) combined with iconic similarity measures [44] could also be an interesting additional component of the registration criterion. Last but not least, domain/problem specific parameter learning [2,24] towards improving the proposed models could have a positive influence on the obtained results.…”

(Hyper)-graphical models in biomedical image analysis

“…In such a context global shape priors could be built through the concatenation of local constraints expressed differently depending on the context [27], resulting on a powerful and flexible model both in terms of learning as well as in terms of inference using generic graph-based optimization methods. A different approach was adopted in [28,29] where the prior knowledge is expressed globally through a multi-class statistical atlas and encoded as constraint throughout the random walker algorithm [24]. Furthermore, in order to define the optimal tradeoff between data, prior and smoothness constraints a mathematically elegant and principled learning framework was introduced in [29].…”

“…A different approach was adopted in [28,29] where the prior knowledge is expressed globally through a multi-class statistical atlas and encoded as constraint throughout the random walker algorithm [24]. Furthermore, in order to define the optimal tradeoff between data, prior and smoothness constraints a mathematically elegant and principled learning framework was introduced in [29]. The work presented in [30] bears concept similarities with the one in [28].…”

Discrete Visual Perception

“…Later the idea of introducing prior knowledge [38], [40] in this context was studied using statistical models of varying complexity, like mono-modal distributions or linear sub-spaces per class. Multi-class segmentation of striated Muscles in NMR Images was the clinical case being investigated in this context where the problem of optimal parameter setting was also investigated through machine learning methods [41]. Continuous methods have certain strengths but also exhibit a number of limitations in particular during the inference process like for example their strong dependency from the initial conditions.…”