Deep Convolutional Encoder Networks for Multiple Sclerosis Lesion Segmentation
We propose a novel segmentation approach based on deep convolutional encoder networks and apply it to the segmentation of multiple sclerosis (MS) lesions in magnetic resonance images. Our model is a neural network that has both convolutional and deconvolu
- PDF / 677,186 Bytes
- 9 Pages / 439.363 x 666.131 pts Page_size
- 16 Downloads / 201 Views
4
Department of Electrical and Computer Engineering, UBC 2 Department of Radiology, UBC 3 Division of Neurology, UBC MS/MRI Research Group, University of British Columbia, Vancouver, Canada
Abstract. We propose a novel segmentation approach based on deep convolutional encoder networks and apply it to the segmentation of multiple sclerosis (MS) lesions in magnetic resonance images. Our model is a neural network that has both convolutional and deconvolutional layers, and combines feature extraction and segmentation prediction in a single model. The joint training of the feature extraction and prediction layers allows the model to automatically learn features that are optimized for accuracy for any given combination of image types. In contrast to existing automatic feature learning approaches, which are typically patch-based, our model learns features from entire images, which eliminates patch selection and redundant calculations at the overlap of neighboring patches and thereby speeds up the training. Our network also uses a novel objective function that works well for segmenting underrepresented classes, such as MS lesions. We have evaluated our method on the publicly available labeled cases from the MS lesion segmentation challenge 2008 data set, showing that our method performs comparably to the state-of-theart. In addition, we have evaluated our method on the images of 500 subjects from an MS clinical trial and varied the number of training samples from 5 to 250 to show that the segmentation performance can be greatly improved by having a representative data set. Keywords: Segmentation, multiple sclerosis lesions, MRI, machine learning, unbalanced classification, deep learning, convolutional neural nets.
1
Introduction
Multiple sclerosis (MS) is an inflammatory and demyelinating disease of the central nervous system, and is characterized by the formation of lesions, primarily visible in the white matter on conventional magnetic resonance images (MRIs). Imaging biomarkers based on the delineation of lesions, such as lesion load and lesion count, have established their importance for assessing disease progression and treatment effect. However, lesions vary greatly in size, shape, intensity
c Springer International Publishing Switzerland 2015 N. Navab et al. (Eds.): MICCAI 2015, Part III, LNCS 9351, pp. 3–11, 2015. DOI: 10.1007/978-3-319-24574-4_1
4
T. Brosch et al.
and location, which makes their automatic and accurate segmentation challenging. Many automatic methods have been proposed for the segmentation of MS lesions over the last two decades, which can be classified into unsupervised and supervised methods. Unsupervised methods do not require a labeled data set for training. Instead, lesions are identified as an outlier class using, e.g., clustering methods [1] or dictionary learning and sparse coding to model healthy tissue [2]. Current supervised approaches typically start with a large set of features, either predefined by the user [3] or gathered in a feature extraction step, which is followed by a sep
Data Loading...
