Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation
Brain tumors are among the deadliest forms of cancer, characterized by abnormal proliferation of brain cells. While early identification of brain tumors can greatly aid in their therapy, the process of manual segmentation performed by expert doctors, which is often time-consuming, tedious, and prone...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-02-01
|
Series: | Sensors |
Subjects: | |
Online Access: | https://www.mdpi.com/1424-8220/23/4/2346 |
_version_ | 1797618228964360192 |
---|---|
author | Azka Rehman Muhammad Usman Abdullah Shahid Siddique Latif Junaid Qadir |
author_facet | Azka Rehman Muhammad Usman Abdullah Shahid Siddique Latif Junaid Qadir |
author_sort | Azka Rehman |
collection | DOAJ |
description | Brain tumors are among the deadliest forms of cancer, characterized by abnormal proliferation of brain cells. While early identification of brain tumors can greatly aid in their therapy, the process of manual segmentation performed by expert doctors, which is often time-consuming, tedious, and prone to human error, can act as a bottleneck in the diagnostic process. This motivates the development of automated algorithms for brain tumor segmentation. However, accurately segmenting the enhanced and core tumor regions is complicated due to high levels of inter- and intra-tumor heterogeneity in terms of texture, morphology, and shape. This study proposes a fully automatic method called the selective deeply supervised multi-scale attention network (SDS-MSA-Net) for segmenting brain tumor regions using a multi-scale attention network with novel selective deep supervision (SDS) mechanisms for training. The method utilizes a 3D input composed of five consecutive slices, in addition to a 2D slice, to maintain sequential information. The proposed multi-scale architecture includes two encoding units to extract meaningful global and local features from the 3D and 2D inputs, respectively. These coarse features are then passed through attention units to filter out redundant information by assigning lower weights. The refined features are fed into a decoder block, which upscales the features at various levels while learning patterns relevant to all tumor regions. The SDS block is introduced to immediately upscale features from intermediate layers of the decoder, with the aim of producing segmentations of the whole, enhanced, and core tumor regions. The proposed framework was evaluated on the BraTS2020 dataset and showed improved performance in brain tumor region segmentation, particularly in the segmentation of the core and enhancing tumor regions, demonstrating the effectiveness of the proposed approach. Our code is publicly available. |
first_indexed | 2024-03-11T08:10:11Z |
format | Article |
id | doaj.art-c05c132ac7514c1ca1cb0fd5db2af4f6 |
institution | Directory Open Access Journal |
issn | 1424-8220 |
language | English |
last_indexed | 2024-03-11T08:10:11Z |
publishDate | 2023-02-01 |
publisher | MDPI AG |
record_format | Article |
series | Sensors |
spelling | doaj.art-c05c132ac7514c1ca1cb0fd5db2af4f62023-11-16T23:13:34ZengMDPI AGSensors1424-82202023-02-01234234610.3390/s23042346Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor SegmentationAzka Rehman0Muhammad Usman1Abdullah Shahid2Siddique Latif3Junaid Qadir4Center for Artificial Intelligence in Medicine and Imaging, HealthHub Co., Ltd., Seoul 06524, Republic of KoreaDepartment of Computer Science and Engineering, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of KoreaCenter for Artificial Intelligence in Medicine and Imaging, HealthHub Co., Ltd., Seoul 06524, Republic of KoreaFaculty of Health, Engineering and Sciences, University of Southern Queensland, Springfield 4300, AustraliaDepartment of Computer Science and Engineering, College of Engineering, Qatar University, Doha 2713, QatarBrain tumors are among the deadliest forms of cancer, characterized by abnormal proliferation of brain cells. While early identification of brain tumors can greatly aid in their therapy, the process of manual segmentation performed by expert doctors, which is often time-consuming, tedious, and prone to human error, can act as a bottleneck in the diagnostic process. This motivates the development of automated algorithms for brain tumor segmentation. However, accurately segmenting the enhanced and core tumor regions is complicated due to high levels of inter- and intra-tumor heterogeneity in terms of texture, morphology, and shape. This study proposes a fully automatic method called the selective deeply supervised multi-scale attention network (SDS-MSA-Net) for segmenting brain tumor regions using a multi-scale attention network with novel selective deep supervision (SDS) mechanisms for training. The method utilizes a 3D input composed of five consecutive slices, in addition to a 2D slice, to maintain sequential information. The proposed multi-scale architecture includes two encoding units to extract meaningful global and local features from the 3D and 2D inputs, respectively. These coarse features are then passed through attention units to filter out redundant information by assigning lower weights. The refined features are fed into a decoder block, which upscales the features at various levels while learning patterns relevant to all tumor regions. The SDS block is introduced to immediately upscale features from intermediate layers of the decoder, with the aim of producing segmentations of the whole, enhanced, and core tumor regions. The proposed framework was evaluated on the BraTS2020 dataset and showed improved performance in brain tumor region segmentation, particularly in the segmentation of the core and enhancing tumor regions, demonstrating the effectiveness of the proposed approach. Our code is publicly available.https://www.mdpi.com/1424-8220/23/4/2346brain tumor segmentation3D segmentationselective deep supervision |
spellingShingle | Azka Rehman Muhammad Usman Abdullah Shahid Siddique Latif Junaid Qadir Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation Sensors brain tumor segmentation 3D segmentation selective deep supervision |
title | Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation |
title_full | Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation |
title_fullStr | Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation |
title_full_unstemmed | Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation |
title_short | Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation |
title_sort | selective deeply supervised multi scale attention network for brain tumor segmentation |
topic | brain tumor segmentation 3D segmentation selective deep supervision |
url | https://www.mdpi.com/1424-8220/23/4/2346 |
work_keys_str_mv | AT azkarehman selectivedeeplysupervisedmultiscaleattentionnetworkforbraintumorsegmentation AT muhammadusman selectivedeeplysupervisedmultiscaleattentionnetworkforbraintumorsegmentation AT abdullahshahid selectivedeeplysupervisedmultiscaleattentionnetworkforbraintumorsegmentation AT siddiquelatif selectivedeeplysupervisedmultiscaleattentionnetworkforbraintumorsegmentation AT junaidqadir selectivedeeplysupervisedmultiscaleattentionnetworkforbraintumorsegmentation |