Research on ME- NBI Gastric Lesion Recognition System based on Improved UNet Structure

Qiu, Kaili; Bu, Xiongzhu; Zhou, Hui; Liu, Leheng

doi:10.1109/iwecai55315.2022.00099

Cited by 6 publications

(5 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Standard, well-known data augmentation methods were used by five works considered in this review, including image rotation [29]- [33], image flipping [33], image cropping [29], [32], image inversion [31], color transformation and noise addition [31]. A common approach observed by Itoh et al [30], Yan et al [32], and K. Qiu et al [33] was the use of transfer learning along with fine-tuning techniques to increase the model's performance and dispute the lack of available data. A pretrained model was obtained using a large dataset of ImageNet natural images and freezing the shallow network for fine-tuning.…”

Section: Algorithmsmentioning

confidence: 99%

“…By adopting this method, the authors sought to Different methods of dataset preprocessing were observed in the papers reviewed. Some of the dataset preparations for the training phase included the cropping of redundant image parts [33], for example, black frames of the selected original images for the development of the ID system [32]. In a different work, realized by Hatami et al [29], the images available on the dataset presented a size of 460x475, and resources to the ROI and processes such as crop and rotation by image preprocessing experts ended up reaching 3673 images with a size of 32x32.…”

Section: Algorithmsmentioning

confidence: 99%

“…They all included upper endoscopic images focused on the stomach and obtained written consent from patients or protected their identities for ethical reasons. While some works prioritized obtaining high-quality images only [33], others focused on the specific endoscopic technique used for acquiring the images [37]. In Yan et al…”

Section: Datasetsmentioning

confidence: 99%

“…In the works reviewed, the datasets were composed of different image modalities, where some considered only white light endoscopy (WLE) frames [37], [39], [30] while others considered narrow-band imaging (NBI) and magnifying endoscopy with NBI (ME-NBI) data [32], [33]. Furthermore, three need to define from what endoscopic technique the dataset was obtained [29], [31], [40].…”

Section: Datasetsmentioning

confidence: 99%

See 3 more Smart Citations

Generative Adversarial Networks for Augmenting Endoscopy Image Datasets of Stomach Precancerous Lesions: A Review

Magalhães,

Neto,

Cunha

2023

IEEE Access

View full text Add to dashboard Cite

Gastric cancer (GC) is still a significant public health issue, among the most common and deadly cancers globally. The identification and characterization of precancerous lesions of the stomach using endoscopy are crucial for determining the risk of cancer and guiding appropriate surveillance. In this scenario, deep learning (DL)-based computer vision methods have the potential to help us classify and identify particular patterns in endoscopic images, leading to a more accurate classification of these types of lesions. The quantity and quality of the data used heavily influence the classification performance of DL networks. However, one of the major setbacks for developing high-performance DL classification models is the typical need for more available data in the medical field. This review explores the use of Generative Adversarial Networks (GANs) and classical data augmentation techniques for improving the classification of precancerous stomach lesions. GANs are DL models that have shown promising results in generating synthetic data, which can be used to augment limited medical datasets. This review discusses recent studies that have implemented GANs and classical data augmentation methods to improve the accuracy of cancerous lesion classification. The results indicate that GANs can effectively increase the dataset's size, enhance the classification models' performance, and, in some cases, obtain superior results than classical data augmentation. Furthermore, this review highlights the challenges and limitations of the recent works using GANs and classical data augmentation techniques in medical imaging analysis and proposes directions for future research.

show abstract

Section: Algorithmsmentioning

confidence: 99%

Section: Algorithmsmentioning

confidence: 99%

Section: Datasetsmentioning

confidence: 99%

Section: Datasetsmentioning

confidence: 99%

See 2 more Smart Citations

Generative Adversarial Networks for Augmenting Endoscopy Image Datasets of Stomach Precancerous Lesions: A Review

Magalhães,

Neto,

Cunha

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…All previous works on GIM classification [15][16][17][18][19][20][21] and segmentation [22][23][24][25] were unable to attain real-time inference speed. In practical use, such studies could not be implemented.…”

Section: List Of Tablesmentioning

confidence: 99%

Real-time gastric intestinal metaplasia semantic segmentation with multiple abnormalities using deep learning approach

Pornvoraphat

View full text Add to dashboard Cite

This thesis declares the segmentation of gastric intestinal metaplasia (GIM) in real-time. Recently, GIM segmentation of endoscopic images has been conducted to distinguish GIM from a healthy stomach. However, achieving real-time detection is difficult. Challenging conditions include multiple color modes (white light endoscopy and narrow-band imaging), other abnormal lesions (erosion and ulcer), noisy labels, etc. Herein, our model is based on BiSeNet and can overcome the many issues regarding GIM. Applying auxiliary head and loss boosts the performance on multiple color modes. In addition, pre-processing techniques, including location-wise negative sampling, jigsaw augmentation, and label smoothing, are utilized to improve detection performance. Finally, the decision threshold can be independently altered for each color mode. Work undertaken at King Chulalongkorn Memorial Hospital examined 940 histologically proven GIM images and 1239 non-GIM images, obtained over 173 FPS. In terms of accuracy, our model outperforms all baselines. Our results reveal F1-score, sensitivity, specificity, accuracy, and mean intersection over union (IoU), achieving 91%, 91%, 96%, 94%, and 55%, respectively. In addition, the effectiveness of the proposed methods was validated on baseline models, achieving F1-score and IoU values of 93% and 56% for STDC2-Seg50 and 93% and 56% for BlazeNeo.

show abstract

Letter to the Editor: Comment on: Chromoendoscopy in Combination with Random Biopsies for Patients with Pathogenic CDH1 Mutations Undergoing Endoscopic Surveillance

Hojati

Asl

2022

J Gastrointest Canc

View full text Add to dashboard Cite

Research on ME- NBI Gastric Lesion Recognition System based on Improved UNet Structure

Cited by 6 publications

References 2 publications

Generative Adversarial Networks for Augmenting Endoscopy Image Datasets of Stomach Precancerous Lesions: A Review

Generative Adversarial Networks for Augmenting Endoscopy Image Datasets of Stomach Precancerous Lesions: A Review

Real-time gastric intestinal metaplasia semantic segmentation with multiple abnormalities using deep learning approach

Letter to the Editor: Comment on: Chromoendoscopy in Combination with Random Biopsies for Patients with Pathogenic CDH1 Mutations Undergoing Endoscopic Surveillance

Contact Info

Product

Resources

About