Medical Image Super-Resolution Based on the Generative Adversarial Network

Huang, Xu; Zhang, Qianyao; Wang, Guoli; Guo, Xuemei; Li, Zhonghua

doi:10.1007/978-981-32-9686-2_29

Cited by 7 publications

(2 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This situation can cause overfitting in the network structure. In order to prevent this situation and increase the training performance, normalization layers can be added to the network structure (Huang et al 2019). The main task of the normalization layer is to bring the values formed as a result of multiplications to a specific range and transmit the appropriate values to the next layer.…”

Section: Proposed Methodsmentioning

confidence: 99%

CNN-Based Approach for Overlapping Erythrocyte Counting and Cell Type Classification in Peripheral Blood Images

Pala¹,

Çimen

Yıldız

et al. 2022

Chaos Theory and Applications

View full text Add to dashboard Cite

Automatic analysis of cell numbers and types from blood smear images is essential for diagnosing and treating many diseases. Peripheral smear has been used for many years and is a gold standard method. However, the overlap in cells during the peripheral smear process may cause incorrectly predicted results in counting blood cells and classifying cell types. This problem can occur both in automated systems and in manual inspections by experts. Convolutional neural networks provide reliable results for segmentation and classification problems in the medical field. However, creating ground truth labels in the data during the segmentation process is a time-consuming and error-prone process. This study proposes a new CNN-based strategy to eliminate the overlap-induced counting problem in peripheral smear blood samples and to determine the blood cell type with high accuracy. In the proposed method, images of the entire slide were divided into sub-images, block by block, using adaptive image processing techniques to identify the overlapping cells and cell types. CNN was used to classify the number of cells separated from the original images into sub-images by blocks. The proposed method both counts overlapping red blood cells and distinguishes RBC-WBC with an accuracy rate of 99.73%. The results show that the proposed method can be adapted to areas where high-resolution images are found and reliable results.

show abstract

Section: Proposed Methodsmentioning

confidence: 99%

CNN-Based Approach for Overlapping Erythrocyte Counting and Cell Type Classification in Peripheral Blood Images

Pala¹,

Çimen

Yıldız

et al. 2022

Chaos Theory and Applications

View full text Add to dashboard Cite

show abstract

“…Moreover, this model was enhanced by the researchers. Thus, by using a feature similarity index for image quality assessment (FISM) instead of original mean square error (MSE), Huang et al, [28] improved the SRGAN model in terms of perceptual loss and increased the histopathology image resolution. Also, SRGAN-SQE was proposed by Upadhyay and Awate [29] by adding an autoencoder in the pre-processing phase to intensify the resolution of the breast cancer histopathology images and employing a heavy-tailed non-Gaussian distribution probability density loss function on the residuals.…”

Section: Srgan For Medical Histopathology Breast Cancer Imagingmentioning

confidence: 99%

Breast Cancer Histopathology Image Super-Resolution Using Wide-Attention GAN With Improved Wasserstein Gradient Penalty and Perceptual Loss

Shahidi

2021

IEEE Access

View full text Add to dashboard Cite

In the realm of image processing, enhancing the quality of the images is known as a superresolution problem (SR). Among SR methods, a super-resolution generative adversarial network, or SRGAN, has been introduced to generate SR images from low-resolution images. As it is of the utmost importance to keep the size and the shape of the images, while enlarging the medical images, we propose a novel superresolution model with a generative adversarial network to generate SR images with finer details and higher quality to encourage less blurring. By widening residual blocks and using a self-attention layer, our model becomes robust and generalizable as it is able to extract the most important part of the images before upsampling. We named our proposed model as wide-attention SRGAN (WA-SRGAN). Moreover, we have applied improved Wasserstein with a Gradient penalty to stabilize the model while training. To train our model, we have applied images from Camylon 16 database and enlarged them by 2×, 4×, 8×, and 16× upscale factors with the ground truth of the size of 256×256×3. Furthermore, two normalization methods, including batch normalization, and weight normalization have been applied and we observed that weight normalization is an enabling factor to improve metric performance in terms of SSIM. Moreover, several evaluation metrics, such as PSNR, MSE, SSIM, MS-SSIM, and QILV have been applied for having a comprehensive objective comparison with other methods, including SRGAN, A-SRGAN, and bicubial. Also, we performed the job of classification by using a deep learning model called ResNeXt-101 (32x8d) for super-resolution, highresolution, and low-resolution images and compared the outcomes in terms of accuracy score. Finally, the results on breast cancer histopathology images show the superiority of our model by using weight normalization and a batch size of one in terms of restoration of the color and the texture details.

show abstract

An Unsupervised Framework for Joint MRI Super Resolution and Gibbs Artifact Removal

Liu¹,

Chen²,

Chen³

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Medical Image Super-Resolution Based on the Generative Adversarial Network

Cited by 7 publications

References 19 publications

CNN-Based Approach for Overlapping Erythrocyte Counting and Cell Type Classification in Peripheral Blood Images

CNN-Based Approach for Overlapping Erythrocyte Counting and Cell Type Classification in Peripheral Blood Images

Breast Cancer Histopathology Image Super-Resolution Using Wide-Attention GAN With Improved Wasserstein Gradient Penalty and Perceptual Loss

An Unsupervised Framework for Joint MRI Super Resolution and Gibbs Artifact Removal

Contact Info

Product

Resources

About