Image classification-based brain tumour tissue segmentation

  • PDF / 1,751,899 Bytes
  • 16 Pages / 439.642 x 666.49 pts Page_size
  • 82 Downloads / 273 Views

DOWNLOAD

REPORT


Image classification-based brain tumour tissue segmentation Salma Al-qazzaz1,2 · Xianfang Sun3 · Hong Yang4 · Yingxia Yang5 · Ronghua Xu6 · Len Nokes1 · Xin Yang1 Received: 25 October 2019 / Revised: 19 June 2020 / Accepted: 18 August 2020 / © The Author(s) 2020

Abstract Brain tumour tissue segmentation is essential for clinical decision making. While manual segmentation is time consuming, tedious, and subjective, it is very challenging to develop automatic segmentation methods. Deep learning with convolutional neural network (CNN) architecture has consistently outperformed previous methods on such challenging tasks. However, the local dependencies of pixel classes cannot be fully reflected in the CNN models. In contrast, hand-crafted features such as histogram-based texture features provide robust feature descriptors of local pixel dependencies. In this paper, a classification-based method for automatic brain tumour tissue segmentation is proposed using combined CNNbased and hand-crafted features. The CIFAR network is modified to extract CNN-based features, and histogram-based texture features are fused to compensate the limitation in the CIFAR network. These features together with the pixel intensities of the original MRI images are sent to a decision tree for classifying the MRI image voxels into different types of tumour tissues. The method is evaluated on the BraTS 2017 dataset. Experiments show that the proposed method produces promising segmentation results. Keywords Brain tumour segmentation · Multi-modal MRI · Convolutional neural networks · Decision tree

1 Introduction Brain gliomas are tumours associated with glial cells. They are primary tumour among adults. Investigations on improving diagnosis of gliomas are thus highly desirable [23, 26]. In the medical imaging field, the decision tree (DT) classifier has been chosen as the best classifier among other conventional classifiers for brain tumour segmentation [22, 31]. Some hand-designed features were exploited as input to the DT classifier [6, 28, 32]. In order to provide good description of different tissue classes, a large number of features are needed. However, a high dimensional feature vector will cause the problem of high  Xin Yang

[email protected]

Extended author information available on the last page of the article.

Multimedia Tools and Applications

complexity with high computational cost. This limitation was solved using another variant of discriminating approaches. Recently, deep convolutional neural network (CNN) methods have attracted great attention in the image classification field. Some of the CNN-based approaches have been adopted for segmentation of brain tumour in medical images, especially in MRI data [2, 12, 27, 39]. The classification in the CNN-based approaches is performed on pixels. Therefore, the main limitation in CNN-based methods is that the local pixel dependencies are not fully considered for pixel classification [24, 37, 38]. On the other hand, some hand-crafted features such as histogram-based texture features