A fusion of salient and convolutional features applying healthy templates for MRI brain tumor segmentation
- PDF / 2,048,243 Bytes
- 18 Pages / 439.642 x 666.49 pts Page_size
- 67 Downloads / 170 Views
A fusion of salient and convolutional features applying healthy templates for MRI brain tumor segmentation ´ 1,2 · Levente Kovacs ´ 1 · Andrea Manno-Kovacs1,2 Petra Takacs Received: 2 March 2019 / Revised: 24 February 2020 / Accepted: 15 September 2020 / © The Author(s) 2020
Abstract This paper proposes an improved brain tumor segmentation method based on visual saliency features on MRI image volumes. The proposed method introduces a novel combination of multiple MRI modalities used as pseudo-color channels for highlighting the potential tumors. The novel pseudo-color model incorporates healthy templates generated from the MRI slices without tumors. The constructed healthy templates are also used during the training of neural network models. Based on a saliency map built using the pseudo-color templates, combination models are proposed, fusing the saliency map with convolutional neural networks’ prediction maps to improve predictions and to reduce the networks’ eventual overfitting which may result in weaker predictions for previously unseen cases. By introducing the combination technique for deep learning techniques and saliency-based, handcrafted feature models, the fusion approach shows good abstraction capabilities and it is able to handle diverse cases that the networks were less trained for. The proposed methods were tested on the BRATS2015 and BRATS2018 databases, and the quantitative results show that hybrid models (including both trained and handcrafted features) can be promising alternatives for reaching higher segmentation performance. Moreover, healthy templates can provide additional information for the training process, enhancing the prediction performance of neural network models. Keywords Visual saliency · Medical image segmentation · Brain tumor detection · Convolutional neural networks
Andrea Manno-Kovacs
[email protected] Petra Tak´acs [email protected] Levente Kov´acs [email protected] 1
Institute for Computer Science and Control SZTAKI, Budapest, Hungary
2
Faculty of Information Technology and Bionics P´azm´any P´eter Catholic University, Budapest, Hungary
Multimedia Tools and Applications
1 Introduction In the last decade, cancer became one of the leading causes of deaths in higher income countries. The earlier the disease is diagnosed, the higher the chance that the patient can be successfully treated. Therefore, quantitative imaging techniques, such as computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) play a dominant role in early diagnosis. In the last few years, with the significant improvement of these non-invasive techniques, the emphasis has shifted to the efficient processing of the diverse data. Gliomas are frequent primary brain tumors in adults [9]. Being highly malignant, this type covers a large portion of all malignant brain tumors. In case of patients with such brain tumors, the role of non-invasive imaging techniques is even more important, as repeated tumor biopsies have a high associated risk. The
Data Loading...
