Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images
Background and purpose: MRI images timely and accurately reflect ischemic injuries to the brain tissues and, therefore, can support clinical decision-making of acute ischemic stroke (AIS). To maximize the information provided by the MRI images, we leverage deep learning models to segment, classify,...
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2022-01-01
|
| Series: | NeuroImage: Clinical |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2213158222001097 |
| _version_ | 1828348231605551104 |
|---|---|
| author | Yi-Chia Wei Wen-Yi Huang Chih-Yu Jian Chih-Chin Heather Hsu Chih-Chung Hsu Ching-Po Lin Chi-Tung Cheng Yao-Liang Chen Hung-Yu Wei Kuan-Fu Chen |
| author_facet | Yi-Chia Wei Wen-Yi Huang Chih-Yu Jian Chih-Chin Heather Hsu Chih-Chung Hsu Ching-Po Lin Chi-Tung Cheng Yao-Liang Chen Hung-Yu Wei Kuan-Fu Chen |
| author_sort | Yi-Chia Wei |
| collection | DOAJ |
| description | Background and purpose: MRI images timely and accurately reflect ischemic injuries to the brain tissues and, therefore, can support clinical decision-making of acute ischemic stroke (AIS). To maximize the information provided by the MRI images, we leverage deep learning models to segment, classify, and map lesion distributions of AIS. Methods: We evaluated brain MRI images of AIS patients from 2017 to 2020 at a tertiary teaching hospital and developed the Semantic Segmentation Guided Detector Network (SGD-Net), composed of the first U-shaped model for segmentation in diffusion-weighted imaging (DWI) and the second model for binary classification of lesion size (lacune vs. non-lacune) and circulatory territory of lesion location (anterior vs. posterior circulation). Next, we modified the two-stage deep learning model into SGD-Net Plus by automatically segmenting AIS lesions in DWI images and registering the lesion in T1-weighted images and the brain atlases. Results: The final enrollment (216 patients with 4606 slices) was divided into 80% for model development and 20% for testing. S1 model segmented AIS lesions in DWI images accurately with a pixel accuracy > 99% (Dice 0.806–0.828 and IoU 0.675–707). In comprehensive evaluation of classification performance, the two-stage SGD-Net outperformed the traditional one-stage models in classifying AIS lesion size (accuracy 0.867–0.956 vs. 0.511–0.867, AUROC 0.962–0.992 vs. 0.528–0.937, AUPRC 0.964–0.994 vs. 0.549–0.938) and location (accuracy 0.860–0.930 vs. 0.326–0.721, AUROC 0.936–0.988 vs. 0.493–0.833, AUPRC 0.883–0.978 vs. 0.365–0.695). The precise lesion segmentation at the first stage of the deep learning model was the basis for further application. After that, the modified two-stage model SGD-Net Plus accurately reported the volume, region percentage, and lesion percentage of each region on the selected brain atlas. Its reports provided clear descriptions and quantifications of the AIS-related brain injuries on white matter tracts, Brodmann areas, and cytoarchitectonic areas. Conclusion: Domain knowledge-oriented design of artificial intelligence applications can deepen our understanding of patients’ conditions and strengthen the use of MRI for patient care. SGD-Net precisely segments AIS lesions on DWI and accurately classifies the lesions. In addition, SGD-Net Plus maps the AIS lesions and quantifies their occupancy in each brain region. They are practical tools to meet the clinical needs and enrich educational resources of neuroimage. |
| first_indexed | 2024-04-14T00:51:53Z |
| format | Article |
| id | doaj.art-8f261e5e6879485e93f6080e1e2e0534 |
| institution | Directory Open Access Journal |
| issn | 2213-1582 |
| language | English |
| last_indexed | 2024-04-14T00:51:53Z |
| publishDate | 2022-01-01 |
| publisher | Elsevier |
| record_format | Article |
| series | NeuroImage: Clinical |
| spelling | doaj.art-8f261e5e6879485e93f6080e1e2e05342022-12-22T02:21:46ZengElsevierNeuroImage: Clinical2213-15822022-01-0135103044Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI imagesYi-Chia Wei0Wen-Yi Huang1Chih-Yu Jian2Chih-Chin Heather Hsu3Chih-Chung Hsu4Ching-Po Lin5Chi-Tung Cheng6Yao-Liang Chen7Hung-Yu Wei8Kuan-Fu Chen9Department of Neurology, Chang Gung Memorial Hospital, Keelung, Taiwan; Institute of Neuroscience, National Yang Ming Chiao Tung University, Taipei, Taiwan; Community Medicine Research Center, Chang Gung Memorial Hospital, Keelung, TaiwanDepartment of Neurology, Chang Gung Memorial Hospital, Keelung, Taiwan; Community Medicine Research Center, Chang Gung Memorial Hospital, Keelung, Taiwan; College of Medicine, Chang Gung University, Taoyuan, TaiwanClinical Informatics and Medical Statistics Research Center, Chung Gung University, Taoyuan, TaiwanInstitute of Neuroscience, National Yang Ming Chiao Tung University, Taipei, Taiwan; Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, TaiwanInstitute of Data Science, National Cheng Kung University, Tainan, TaiwanInstitute of Neuroscience, National Yang Ming Chiao Tung University, Taipei, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan; Department of Trauma and Emergency Surgery, Chang Gung Memorial Hospital, Linkou, TaiwanDepartment of Radiology, Chang Gung Memorial Hospital, Keelung, Taiwan; Department of Medical Imaging and Radiological Sciences, Chang Gung University, Taoyuan, TaiwanDepartment of Electrical Engineering, National Taiwan University, Taipei, TaiwanCommunity Medicine Research Center, Chang Gung Memorial Hospital, Keelung, Taiwan; Clinical Informatics and Medical Statistics Research Center, Chung Gung University, Taoyuan, Taiwan; Department of Emergency Medicine, Chang Gung Memorial Hospital, Keelung, Taiwan; Corresponding author at: No.5, Fusing St., Gueishan Dist, Taoyuan City 33302, Taiwan, ROC.Background and purpose: MRI images timely and accurately reflect ischemic injuries to the brain tissues and, therefore, can support clinical decision-making of acute ischemic stroke (AIS). To maximize the information provided by the MRI images, we leverage deep learning models to segment, classify, and map lesion distributions of AIS. Methods: We evaluated brain MRI images of AIS patients from 2017 to 2020 at a tertiary teaching hospital and developed the Semantic Segmentation Guided Detector Network (SGD-Net), composed of the first U-shaped model for segmentation in diffusion-weighted imaging (DWI) and the second model for binary classification of lesion size (lacune vs. non-lacune) and circulatory territory of lesion location (anterior vs. posterior circulation). Next, we modified the two-stage deep learning model into SGD-Net Plus by automatically segmenting AIS lesions in DWI images and registering the lesion in T1-weighted images and the brain atlases. Results: The final enrollment (216 patients with 4606 slices) was divided into 80% for model development and 20% for testing. S1 model segmented AIS lesions in DWI images accurately with a pixel accuracy > 99% (Dice 0.806–0.828 and IoU 0.675–707). In comprehensive evaluation of classification performance, the two-stage SGD-Net outperformed the traditional one-stage models in classifying AIS lesion size (accuracy 0.867–0.956 vs. 0.511–0.867, AUROC 0.962–0.992 vs. 0.528–0.937, AUPRC 0.964–0.994 vs. 0.549–0.938) and location (accuracy 0.860–0.930 vs. 0.326–0.721, AUROC 0.936–0.988 vs. 0.493–0.833, AUPRC 0.883–0.978 vs. 0.365–0.695). The precise lesion segmentation at the first stage of the deep learning model was the basis for further application. After that, the modified two-stage model SGD-Net Plus accurately reported the volume, region percentage, and lesion percentage of each region on the selected brain atlas. Its reports provided clear descriptions and quantifications of the AIS-related brain injuries on white matter tracts, Brodmann areas, and cytoarchitectonic areas. Conclusion: Domain knowledge-oriented design of artificial intelligence applications can deepen our understanding of patients’ conditions and strengthen the use of MRI for patient care. SGD-Net precisely segments AIS lesions on DWI and accurately classifies the lesions. In addition, SGD-Net Plus maps the AIS lesions and quantifies their occupancy in each brain region. They are practical tools to meet the clinical needs and enrich educational resources of neuroimage.http://www.sciencedirect.com/science/article/pii/S2213158222001097SGD-netSGD-net plusAcute ischemic strokeDiffusion-weighted imagingJoint segmentation and classificationLesion distribution and mapping |
| spellingShingle | Yi-Chia Wei Wen-Yi Huang Chih-Yu Jian Chih-Chin Heather Hsu Chih-Chung Hsu Ching-Po Lin Chi-Tung Cheng Yao-Liang Chen Hung-Yu Wei Kuan-Fu Chen Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images NeuroImage: Clinical SGD-net SGD-net plus Acute ischemic stroke Diffusion-weighted imaging Joint segmentation and classification Lesion distribution and mapping |
| title | Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images |
| title_full | Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images |
| title_fullStr | Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images |
| title_full_unstemmed | Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images |
| title_short | Semantic segmentation guided detector for segmentation, classification, and lesion mapping of acute ischemic stroke in MRI images |
| title_sort | semantic segmentation guided detector for segmentation classification and lesion mapping of acute ischemic stroke in mri images |
| topic | SGD-net SGD-net plus Acute ischemic stroke Diffusion-weighted imaging Joint segmentation and classification Lesion distribution and mapping |
| url | http://www.sciencedirect.com/science/article/pii/S2213158222001097 |
| work_keys_str_mv | AT yichiawei semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT wenyihuang semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT chihyujian semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT chihchinheatherhsu semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT chihchunghsu semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT chingpolin semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT chitungcheng semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT yaoliangchen semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT hungyuwei semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages AT kuanfuchen semanticsegmentationguideddetectorforsegmentationclassificationandlesionmappingofacuteischemicstrokeinmriimages |