LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules
Numerous people die from lung cancer every year, making it a serious public health issue. Oftentimes, the symptoms of lung cancer manifest only at a later stage, when it is difficult to treat. Pulmonary nodules are commonly found while screening the lungs using a Computed Tomography (CT) scan, and s...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10776983/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850066248546648064 |
|---|---|
| author | P. C. Sarah Prithvika L. Jani Anbarasi |
| author_facet | P. C. Sarah Prithvika L. Jani Anbarasi |
| author_sort | P. C. Sarah Prithvika |
| collection | DOAJ |
| description | Numerous people die from lung cancer every year, making it a serious public health issue. Oftentimes, the symptoms of lung cancer manifest only at a later stage, when it is difficult to treat. Pulmonary nodules are commonly found while screening the lungs using a Computed Tomography (CT) scan, and some of the nodules may be cancerous. So, an efficient automated pulmonary nodule segmentation system is needed to isolate the pulmonary nodules from the scan images. The doctors can track the nodules that are likely to be malignant and provide early treatment if they become cancerous, thereby improving the patient’s chance of survival. The attention mechanism is a technique that is often used in computer vision to enhance the neural network’s performance. LA-ResUNet, a pulmonary nodule segmentation model, built using ResUNet with a linear attention mechanism and the Leaky ReLU activation function is proposed. LA-ResUNet efficiently segments pulmonary nodules, while achieving a linear time and space complexity. By employing residual blocks, it is possible to construct a deep network without facing the vanishing gradient problem. Additionally, it makes deep network training simpler. Skip connections allow for better gradient flow during training and better information flow between layers. Leaky ReLU addresses the dying ReLU scenario, a situation where some neurons cease to learn when the network is being trained. LA-ResUNet was used on the dataset LIDC-IDRI (The Lung Image Database Consortium and Image Database Resource Initiative) and it produced a dice score coefficient (DSC) of 73.11% and Intersection over Union score (IoU) of 60.62%. |
| format | Article |
| id | doaj-art-e44e00bce6834e569bb1041929d502ee |
| institution | DOAJ |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-e44e00bce6834e569bb1041929d502ee2025-08-20T02:48:49ZengIEEEIEEE Access2169-35362024-01-011218289418290710.1109/ACCESS.2024.351085610776983LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary NodulesP. C. Sarah Prithvika0https://orcid.org/0009-0000-9313-4056L. Jani Anbarasi1https://orcid.org/0000-0002-8904-2236School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, IndiaSchool of Computer Science and Engineering, Vellore Institute of Technology, Chennai, IndiaNumerous people die from lung cancer every year, making it a serious public health issue. Oftentimes, the symptoms of lung cancer manifest only at a later stage, when it is difficult to treat. Pulmonary nodules are commonly found while screening the lungs using a Computed Tomography (CT) scan, and some of the nodules may be cancerous. So, an efficient automated pulmonary nodule segmentation system is needed to isolate the pulmonary nodules from the scan images. The doctors can track the nodules that are likely to be malignant and provide early treatment if they become cancerous, thereby improving the patient’s chance of survival. The attention mechanism is a technique that is often used in computer vision to enhance the neural network’s performance. LA-ResUNet, a pulmonary nodule segmentation model, built using ResUNet with a linear attention mechanism and the Leaky ReLU activation function is proposed. LA-ResUNet efficiently segments pulmonary nodules, while achieving a linear time and space complexity. By employing residual blocks, it is possible to construct a deep network without facing the vanishing gradient problem. Additionally, it makes deep network training simpler. Skip connections allow for better gradient flow during training and better information flow between layers. Leaky ReLU addresses the dying ReLU scenario, a situation where some neurons cease to learn when the network is being trained. LA-ResUNet was used on the dataset LIDC-IDRI (The Lung Image Database Consortium and Image Database Resource Initiative) and it produced a dice score coefficient (DSC) of 73.11% and Intersection over Union score (IoU) of 60.62%.https://ieeexplore.ieee.org/document/10776983/Convolutional neural networkencoder-decoderlinear attentionpulmonary nodule segmentationResUNet |
| spellingShingle | P. C. Sarah Prithvika L. Jani Anbarasi LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules IEEE Access Convolutional neural network encoder-decoder linear attention pulmonary nodule segmentation ResUNet |
| title | LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules |
| title_full | LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules |
| title_fullStr | LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules |
| title_full_unstemmed | LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules |
| title_short | LA-ResUNet: An Efficient Linear Attention Mechanism in ResUNet for the Semantic Segmentation of Pulmonary Nodules |
| title_sort | la resunet an efficient linear attention mechanism in resunet for the semantic segmentation of pulmonary nodules |
| topic | Convolutional neural network encoder-decoder linear attention pulmonary nodule segmentation ResUNet |
| url | https://ieeexplore.ieee.org/document/10776983/ |
| work_keys_str_mv | AT pcsarahprithvika laresunetanefficientlinearattentionmechanisminresunetforthesemanticsegmentationofpulmonarynodules AT ljanianbarasi laresunetanefficientlinearattentionmechanisminresunetforthesemanticsegmentationofpulmonarynodules |