{"title":"Attention-Based Neural Network for Cardiac MRI Segmentation: Application to Strain and Volume Computation","authors":"","doi":"10.1016/j.irbm.2024.100850","DOIUrl":null,"url":null,"abstract":"<div><h3>Context</h3><p>Deep learning algorithms have been widely used for cardiac image segmentation. However, most of these architectures rely on convolutions that hardly model long-range dependencies, limiting their ability to extract contextual information. Moreover, the traditional U-net architecture suffers from the difference of semantic information between feature maps of the encoder and decoder (also known as the semantic gap).</p></div><div><h3>Material and method</h3><p>To address this issue, a new network architecture relying on attention mechanism was introduced. Swin Filtering Blocks (SFB), that use Swin Transformer blocks in a cross-attention manner, were added between the encoder and the decoder to filter information coming from the encoder based on the feature map from the decoder. Attention was also employed at the lowest resolution in the form of a transformer layer to increase the receptive field of the network.</p><p>We conducted experiments to assess both generalization capability and to evaluate how training on all frames of the cardiac cycle rather than only the end-diastole and end-systole impacts strain and segmentation performances.</p></div><div><h3>Results and conclusion</h3><p>Visual inspection of feature maps suggested that Swin Filtering Blocks contribute to the reduction of the semantic gap. Performing attention between all patches using a transformer layer brought higher performance than convolutions. Training the model with all phases of the cardiac cycle resulted in slightly more accurate segmentations while leading to a more noticeable improvement for strain estimation. A limited decrease in performance was observed when testing on out-of-distribution data, but the gap widens for the most apical slices.</p></div>","PeriodicalId":14605,"journal":{"name":"Irbm","volume":null,"pages":null},"PeriodicalIF":5.6000,"publicationDate":"2024-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1959031824000319/pdfft?md5=45a62576b482068e95734d0020169441&pid=1-s2.0-S1959031824000319-main.pdf","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Irbm","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1959031824000319","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, BIOMEDICAL","Score":null,"Total":0}
引用次数: 0
Abstract
Context
Deep learning algorithms have been widely used for cardiac image segmentation. However, most of these architectures rely on convolutions that hardly model long-range dependencies, limiting their ability to extract contextual information. Moreover, the traditional U-net architecture suffers from the difference of semantic information between feature maps of the encoder and decoder (also known as the semantic gap).
Material and method
To address this issue, a new network architecture relying on attention mechanism was introduced. Swin Filtering Blocks (SFB), that use Swin Transformer blocks in a cross-attention manner, were added between the encoder and the decoder to filter information coming from the encoder based on the feature map from the decoder. Attention was also employed at the lowest resolution in the form of a transformer layer to increase the receptive field of the network.
We conducted experiments to assess both generalization capability and to evaluate how training on all frames of the cardiac cycle rather than only the end-diastole and end-systole impacts strain and segmentation performances.
Results and conclusion
Visual inspection of feature maps suggested that Swin Filtering Blocks contribute to the reduction of the semantic gap. Performing attention between all patches using a transformer layer brought higher performance than convolutions. Training the model with all phases of the cardiac cycle resulted in slightly more accurate segmentations while leading to a more noticeable improvement for strain estimation. A limited decrease in performance was observed when testing on out-of-distribution data, but the gap widens for the most apical slices.
期刊介绍:
IRBM is the journal of the AGBM (Alliance for engineering in Biology an Medicine / Alliance pour le génie biologique et médical) and the SFGBM (BioMedical Engineering French Society / Société française de génie biologique médical) and the AFIB (French Association of Biomedical Engineers / Association française des ingénieurs biomédicaux).
As a vehicle of information and knowledge in the field of biomedical technologies, IRBM is devoted to fundamental as well as clinical research. Biomedical engineering and use of new technologies are the cornerstones of IRBM, providing authors and users with the latest information. Its six issues per year propose reviews (state-of-the-art and current knowledge), original articles directed at fundamental research and articles focusing on biomedical engineering. All articles are submitted to peer reviewers acting as guarantors for IRBM''s scientific and medical content. The field covered by IRBM includes all the discipline of Biomedical engineering. Thereby, the type of papers published include those that cover the technological and methodological development in:
-Physiological and Biological Signal processing (EEG, MEG, ECG…)-
Medical Image processing-
Biomechanics-
Biomaterials-
Medical Physics-
Biophysics-
Physiological and Biological Sensors-
Information technologies in healthcare-
Disability research-
Computational physiology-
…