Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism
Image quality plays a critical role in medical image analysis, significantly impacting diagnostic outcomes. Sharp and detailed images are essential for accurate diagnoses, but acquiring high-resolution medical images often demands sophisticated and costly equipment. To address this challenge, this s...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/15/2/867 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832589200538468352 |
|---|---|
| author | Dong Yun Lee Jang Yeop Kim Soo Young Cho |
| author_facet | Dong Yun Lee Jang Yeop Kim Soo Young Cho |
| author_sort | Dong Yun Lee |
| collection | DOAJ |
| description | Image quality plays a critical role in medical image analysis, significantly impacting diagnostic outcomes. Sharp and detailed images are essential for accurate diagnoses, but acquiring high-resolution medical images often demands sophisticated and costly equipment. To address this challenge, this study proposes a convolutional neural network (CNN)-based super-resolution architecture, utilizing a melanoma dataset to enhance image resolution through deep learning techniques. The proposed model incorporates a convolutional self-attention block that combines channel and spatial attention to emphasize important image features. Channel attention uses global average pooling and fully connected layers to enhance high-frequency features within channels. Meanwhile, spatial attention applies a single-channel convolution to emphasize high-frequency features in the spatial domain. By integrating various attention blocks, feature extraction is optimized and further expanded through subpixel convolution to produce high-quality super-resolution images. The model uses L1 loss to generate realistic and smooth outputs, outperforming existing deep learning methods in capturing contours and textures. Evaluations with the ISIC 2020 dataset—containing 33126 training and 10982 test images for skin lesion analysis—showed a 1–2% improvement in peak signal-to-noise ratio (PSNR) compared to very deep super-resolution (VDSR) and enhanced deep super-resolution (EDSR) architectures. |
| format | Article |
| id | doaj-art-91ad5cecf2e8462b98c946cc2198d4d5 |
| institution | Kabale University |
| issn | 2076-3417 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj-art-91ad5cecf2e8462b98c946cc2198d4d52025-01-24T13:21:07ZengMDPI AGApplied Sciences2076-34172025-01-0115286710.3390/app15020867Improving Medical Image Quality Using a Super-Resolution Technique with Attention MechanismDong Yun Lee0Jang Yeop Kim1Soo Young Cho2Department of Defense Acquisition Program, Kwangwoon University, Seoul 01897, Republic of KoreaDepartment of Defense Acquisition Program, Kwangwoon University, Seoul 01897, Republic of KoreaGame Contents Department, Kwangwoon University, Seoul 01897, Republic of KoreaImage quality plays a critical role in medical image analysis, significantly impacting diagnostic outcomes. Sharp and detailed images are essential for accurate diagnoses, but acquiring high-resolution medical images often demands sophisticated and costly equipment. To address this challenge, this study proposes a convolutional neural network (CNN)-based super-resolution architecture, utilizing a melanoma dataset to enhance image resolution through deep learning techniques. The proposed model incorporates a convolutional self-attention block that combines channel and spatial attention to emphasize important image features. Channel attention uses global average pooling and fully connected layers to enhance high-frequency features within channels. Meanwhile, spatial attention applies a single-channel convolution to emphasize high-frequency features in the spatial domain. By integrating various attention blocks, feature extraction is optimized and further expanded through subpixel convolution to produce high-quality super-resolution images. The model uses L1 loss to generate realistic and smooth outputs, outperforming existing deep learning methods in capturing contours and textures. Evaluations with the ISIC 2020 dataset—containing 33126 training and 10982 test images for skin lesion analysis—showed a 1–2% improvement in peak signal-to-noise ratio (PSNR) compared to very deep super-resolution (VDSR) and enhanced deep super-resolution (EDSR) architectures.https://www.mdpi.com/2076-3417/15/2/867attention mechanismconvolutional self-attention blockISIC 2020super-resolution |
| spellingShingle | Dong Yun Lee Jang Yeop Kim Soo Young Cho Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism Applied Sciences attention mechanism convolutional self-attention block ISIC 2020 super-resolution |
| title | Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism |
| title_full | Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism |
| title_fullStr | Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism |
| title_full_unstemmed | Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism |
| title_short | Improving Medical Image Quality Using a Super-Resolution Technique with Attention Mechanism |
| title_sort | improving medical image quality using a super resolution technique with attention mechanism |
| topic | attention mechanism convolutional self-attention block ISIC 2020 super-resolution |
| url | https://www.mdpi.com/2076-3417/15/2/867 |
| work_keys_str_mv | AT dongyunlee improvingmedicalimagequalityusingasuperresolutiontechniquewithattentionmechanism AT jangyeopkim improvingmedicalimagequalityusingasuperresolutiontechniquewithattentionmechanism AT sooyoungcho improvingmedicalimagequalityusingasuperresolutiontechniquewithattentionmechanism |