갑상선 초음파 영상의 평활화 알고리즘에 따른 U-Net 기반 학습 모델 평가
Evaluation of U-Net Based Learning Models according to Equalization Algorithm in Thyroid Ultrasound Imaging
2
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
This study aims to evaluate the performance of the U-Net based learning model that may vary depending on the histogram equalization algorithm. The subject of the experiment were 17 radiology students of this college, and 1,727 data sets in which the region of interest was set in the thyroid after acquiring ultrasound image data were used. The training set consisted of 1,383 images, the validation set consisted of 172 and the test data set consisted of 172. The equalization algorithm was divided into Histogram Equalization(HE) and Contrast Limited Adaptive Histogram Equalization(CLAHE), and according to the clip limit, it was divided into CLAHE8-1, CLAHE8-2. CLAHE8-3. Deep Learning was learned through size control, histogram equalization, Z-score normalization, and data augmentation. As a result of the experiment, the Attention U-Net showed the highest performance from CLAHE8-2 to 0.8355, and the U-Net and BSU-Net showed the highest performance from CLAHE8-3 to 0.8303 and 0.8277. In the case of mIoU, the Attention U-Net was 0.7175 in CLAHE8-2, the U-Net was 0.7098 and the BSU-Net was 0.7060 in CLAHE8-3. This study attempted to confirm the effects of U-Net, Attention U-Net, and BSU-Net models when histogram equalization is performed on ultrasound images. The in- crease in Clip Limit can be expected to increase the ROI match with the prediction mask by clarifying the boundaries, which affects the improvement of the contrast of the thyroid area in deep learning model learning, and consequently affects the performance improvement.
Abstract
This study aims to evaluate the performance of the U-Net based learning model that may vary depending on the histogram equalization algorithm. The subject of the experiment were 17 radiology students of this college, and 1,727 data sets in which the region of interest was set in the thyroid after acquiring ultrasound image data were used. The training set consisted of 1,383 images, the validation set consisted of 172 and the test data set consisted of 172. The equalization algorithm was divided into Histogram Equalization(HE) and Contrast Limited Adaptive Histogram Equalization(CLAHE), and according to the clip limit, it was divided into CLAHE8-1, CLAHE8-2. CLAHE8-3. Deep Learning was learned through size control, histogram equalization, Z-score normalization, and data augmentation. As a result of the experiment, the Attention U-Net showed the highest performance from CLAHE8-2 to 0.8355, and the U-Net and BSU-Net showed the highest performance from CLAHE8-3 to 0.8303 and 0.8277. In the case of mIoU, the Attention U-Net was 0.7175 in CLAHE8-2, the U-Net was 0.7098 and the BSU-Net was 0.7060 in CLAHE8-3. This study attempted to confirm the effects of U-Net, Attention U-Net, and BSU-Net models when histogram equalization is performed on ultrasound images. The in- crease in Clip Limit can be expected to increase the ROI match with the prediction mask by clarifying the boundaries, which affects the improvement of the contrast of the thyroid area in deep learning model learning, and consequently affects the performance improvement.
참고문헌 (20)
- Todsen T, "Ultrasound-guided fine-needle aspiration biopsy of thyroid nodules", Head & Neck, 43(3), pp.1009~1013 (2021)[DOI]
- Ma X, "Diagnosis of thyroid nodules based on image enhancement and deep neural networks", Computational Intelligence and Neuroscience, 2022, pp.5582029~ (2022)[DOI]
- Choi YH, "U-net based transfer learning model for lesion area segmentation on the breast ultrasound images by optimization of encoder", pp.2022
- Shin SY, "A study on residual u-net for semantic segmentation based on deep learning", Journal of Digital Convergence, 19(6), pp.251~258 (2021)[DOI]
- Lee J, "Using deep learning for image-based superficial femoral artery disease detection", pp.2022
- Kim YJ, "A performance comparison of histogram equalization algorithms for cervical cancer classification model", Journal of Biomedical Engineering Research, 42(3), pp.80~85 (2021)
- Yoshimi Y, "Image preprocessing with contrast-limited adaptive histogram equalization improves the segmentation performance of deep learning for the articular disk of the temporomandibular joint on magnetic resonance images", Oral Surg Oral Med Oral Pathol Oral Radiol (2023)[DOI]
- Cho YH, "Quality enhancement of medical images by using nonlinear histogram equalization function", Journal of the Korean Society of Industry Convergence, 13(1), pp.23~30 (2010)[DOI]
- Zuiderveld K, "Graphics gems IV" (1994), Academic Press
- Son JW, "Automatic detection system of underground pipe using 3D GPR exploration data and deep convolutional neural networks", Journal of the Korea Society of Computer and Information, 26(2), pp.27~37 (2021)[DOI]
- "Fully convolutional networks for semantic segmentation"[URL]
- Left atrial segmentation technique based on u-net, "Lee JH", Korea Institute of Communication Sciences, 139~140 (2022)
- "U-net:Convolutional networks for biomedical image segmentation" (2015)[URL]
- Shin SY, "Atrous residual u-net for semantic segmentation in street scenes based on deep learning", Journal of Convergence for Information Technology, 11(10), pp.45~52 (2021)[DOI]
- Hwang DH, "SKU-net : Improved U-net using selective kernel convolution for retinal vessel segmentation", Journal of the Korea Society of Computer and Information, 26(4), pp.29~37 (2021)[DOI]
- Kim IK, "Contactless user identification system using multi-channel palm images facilitated by triple attention u-net and CNN classifier ensemble models", Journal of the Korea Society of Computer and Information, 27(3), pp.33~43 (2022)[DOI]
- Shin HS, "Application and evaluation of the attention u-net using UAV imagery for corn cultivation field extraction", Ecology and Resilient Infrastructure, 8(4), pp.253~265 (2021)[DOI]
- Oh JY, "Evaluating usefulness of deep learning based left ventricle segmentation in cardiac gated blood pool scan", Journal of Radiological Science and Technology, 45(2), pp.151~158 (2022)[DOI]
- Hong JY, "Artificial intelligence based medical imaging : An overview", Journal of Radiological Science and Technology, 43(3), pp.195~208 (2020)[DOI]