3D U-Net for Skull Stripping in Brain MRI

Abstract

Skull stripping in brain magnetic resonance imaging (MRI) is an essential step to analyze images of the brain. Although manual segmentation has the highest accuracy, it is a time-consuming task. Therefore, various automatic segmentation algorithms of the brain in MRI have been devised and proposed previously. However, there is still no method that solves the entire brain extraction problem satisfactorily for diverse datasets in a generic and robust way. To address these shortcomings of existing methods, we propose the use of a 3D-UNet for skull stripping in brain MRI. The 3D-UNet was recently proposed and has been widely used for volumetric segmentation in medical images due to its outstanding performance. It is an extended version of the previously proposed 2D-UNet, which is based on a deep learning network, specifically, the convolutional neural network. We evaluated 3D-UNet skull-stripping using a publicly available brain MRI dataset and compared the results with three existing methods (BSE, ROBEX, and Kleesiek’s method; BSE and ROBEX are two conventional methods, and Kleesiek’s method is based on deep learning). The 3D-UNet outperforms two typical methods and shows comparable results with the specific deep learning-based algorithm, exhibiting a mean Dice coefficient of 0.9903, a sensitivity of 0.9853, and a specificity of 0.9953.

Brain extraction on MRI scans in presence of diffuse glioma: Multi-institutional performance evaluation of deep learning methods and robust modality-agnostic training

“…Many automated methods have previously been developed for brain extraction ( Smith, 2002 ; Segonne et al, 2004 ; Eskildsen et al, 2012 ; Doshi et al, 2013 ; Shattuck et al, 2001 ; Iglesias et al, 2011 ). In recent years, several DL-based approaches have also shown promising results in brain extraction ( Hwang et al, 2019 ). Even though these methods are shown to be widely successful on healthy subjects, they tend to be less accurate when evaluated on brain MRI scans with brain tumors.…”

Deep Learning and Big Data in Healthcare: A Double Review for Critical Beginners

“…Examples are implemented in PyTorch [209], a famous open-source DL library for Python. A DL architecture, which is based on the U-Net network [210,211] has been designed: (1) To predict the arterial blood pressure signal by using the plethysmogram signal (PLETH to ABP example) as input; and (2) to predict the plethysmogram signal (ABP to PLETH example) by using the arterial blood pressure signal as input that is the reverse problem. The architecture is a one-dimensional CNN, called a fully convolutional network, with a deeply-supervised encoder-decoder connected through a series of skip pathways, and in which the input and the output are the same size.…”

Conventional and Deep Learning Methods for Skull Stripping in Brain MRI

“…Each of the aforementioned skull stripping algorithms for brain MRIs possesses pros and cons that alter with the image characteristics; scanning protocol, such as image signal-to-noise ratio, contrast, and resolution; and subject-specific characteristics, such as age and atrophy [91,92]. Furthermore, the algorithms can also deviate in their accuracy in different brain anatomic regions [93]. The development of a hybrid algorithm that intelligently exploits the advantages of the contributing sub-algorithms should attain brain extraction results that are, typically, superior to any individual method.…”

“…Hwang et al [93] suggested that 3D-UNet, the general 3D segmentation network, could be used for skull-stripping problems. The authors showed that the performance is comparable to the method developed by Kleesiek et al [130].…”