Search

Epilepsy is a prevalent neurological disorder that threatens human health in the world. The most commonly used method to detect epilepsy is using the electroencephalogram (EEG). However, epilepsy detection from the EEG is time-consuming and error-prone work because of the varying levels of experience we find in physicians. To tackle this challenge, in this paper, we propose a multi-scale non-local (MNL) network to achieve automatic EEG signal detection. Our MNL-Network is based on 1D convolution neural network involving two specific layers to improve the classification performance. One layer is named the signal pooling layer which incorporates three different sizes of 1D max-pooling layers to learn the multi-scale features from the EEG signal. The other one is called a multi-scale non-local layer, which calculates the correlation of different multi-scale extracted features and outputs the correlative encoded features to further enhance the classification performance. To evaluate the effectiveness of our model, we conduct experiments on the Bonn dataset. The experimental results demonstrate that our MNL-Network could achieve competitive results in the EEG classification task.

Epilepsy is a prevalent neurological disorder that threatens human health in the world. The most commonly used method to detect epilepsy is using the electroencephalogram (EEG). However, epilepsy detection from the EEG is time-consuming and error-prone work because of the varying levels of experience we find in physicians. To tackle this challenge, in this paper, we propose a multi-scale non-local (MNL) network to achieve automatic EEG signal detection. Our MNL-Network is based on 1D convolution neural network involving two specific layers to improve the classification performance. One layer is named the signal pooling layer which incorporates three different sizes of 1D max-pooling layers to learn the multi-scale features from the EEG signal. The other one is called a multi-scale non-local layer, which calculates the correlation of different multi-scale extracted features and outputs the correlative encoded features to further enhance the classification performance. To evaluate the effectiveness of our model, we conduct experiments on the Bonn dataset. The experimental results demonstrate that our MNL-Network could achieve competitive results in the EEG classification task. [ABSTRACT FROM AUTHOR]

In medical imaging, multiple sclerosis (MS) lesions can lead to confounding effects in automatic morphometric processing tools such as registration, segmentation and cortical extraction and subsequently alter individual longitudinal measurements. Multiple magnetic resonance imaging (MRI) inpainting techniques have been proposed to decrease the impact of MS lesions in medical image processing, however, most of these methods make the assumption that lesions only affect white matter. Here, we propose a method to fill lesion regions using the patch-based non-local mean (NLM) strategy. The method consists of a hierarchical concentric filling strategy after identification of the lesion region. The lesion is filled iteratively, based on the surrounding tissue intensity, using an onion peel strategy. This concentric technique presents the advantage of preserving the local information and therefore the continuity of the anatomy and does not require identification of any a priori normal brain tissues. The method is first evaluated on simulated artificial MS lesions and, second, we performed a power analysis to evaluate brain atrophy and ventricular growth in patients with MS. The method was compared to two different publicly available methods (FSL lesion fill and Lesion LEAP) and more classic method approaches. The proposed method was shown to exceed the other methods in reproducing the fidelity of the images where the lesions were inpainted. The method also improved the power to detect brain atrophy or ventricular growth by decreasing the sample size by 25% in the presence of MS lesions.

In medical imaging, multiple sclerosis (MS) lesions can lead to confounding effects in automatic morphometric processing tools such as registration, segmentation and cortical extraction, and subsequently alter individual longitudinal measurements. Multiple magnetic resonance imaging (MRI) inpainting techniques have been proposed to decrease the impact of MS lesions in medical image processing, however, most of these methods make the assumption that lesions only affect white matter. Here, we propose a method to fill lesion regions using the patch-based non-local mean (NLM) strategy. The method consists of a hierarchical concentric filling strategy after identification of the lesion region. The lesion is filled iteratively, based on the surrounding tissue intensity, using an onion peel strategy. This concentric technique presents the advantage of preserving the local information and therefore the continuity of the anatomy and does not require identification of any a priori normal brain tissues. The method is first evaluated on 20 healthy subjects with simulated artificial MS lesions where we assessed our technique by measuring the peak signal-to-noise ratio (PSNR) of the images with inpainted lesion and the original healthy images. Second, in order to assess the impact of lesion filling on longitudinal image analyses, we performed a power analysis with sample size estimation to evaluate brain atrophy and ventricular growth in patients with MS. The method was compared to two different publicly available methods (FSL lesion fill and Lesion LEAP) and a more classic method, which fills the region with intensities similar to that of the surrounding healthy white matter tissue or mask the lesions. The proposed method was shown to exceed the other methods in reproducing the fidelity of healthy subject images where the lesions were inpainted. The method also improved the power to detect brain atrophy or ventricular growth by decreasing the sample size by 25% in the presence of MS lesions. [ABSTRACT FROM AUTHOR]