Use of U-Net Convolutional Neural Networks for Automated Segmentation of Fecal Material for Objective Evaluation of Bowel Preparation Quality in Colonoscopy

Wang, Yen‐Po; Jheng, Ying Chun; Sung, Kuang-Yi; Lin, Horng-Horng; Hsin, I-Fang; Chen, Ping‐Hsien; Lu, David; Wang, Yuan‐Jen; Hou, Ming‐Chih; Lee, Fa-Yauh; Lu, Chengqiang

doi:10.3390/diagnostics12030613

Cited by 10 publications

(6 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The same reason holds for the pulmonary FoV; (iii) All UNet classes had four to five layers, expect sUNet that had up to a maximum of 13 layers [42][43][44][45][46][47][48][49][50][51][52][53][54]. Note that as the number of layers increase, the DL system becomes more complex; (iv) Cross-Entropy (CE) loss function was most common or popular in all the five types of UNet , while, Dice loss function was also part of cUNet and sUNet classes ; (v) sUNet and acUNet were the two sets of classes which embraced multicenter studies [49,.…”

Section: B Five Types Of Unet and Their Attributesmentioning

confidence: 99%

“…To begin with, the encoder is the most adapted and most changeable component of the UNet architecture. Since it is practically not possible to study each of the architectural variations in the encoder, we have therefore listed here the 23 variations (E1 to E23, representing encoder changes) along with their references in a tabular format and it is as follows: (E1) conventional system (Ronneberger) [43][44][45][46][47][48][49][50][51][52]90]; (E2) cascade of convolutions [77,91,99,116,117]; (E3) parallel convolutions (multiple convolution network) [57]; (E4) convolution with dropout [70,76,86,95,101,102,134,138]; (E5) Residual network [76,78,105,129,135,138,[149][150][151]; (E6) Xception encoder [56,88,112]; (E7) encoder layers with independent inputs [104,140]; (E8) squeeze excitation (SE) network [92,103,138]; (E9) pooling types (max pooling, global average pooling) [95]; (E10) input image dimension change with changing filter (channe...…”

Section: A Encoder Variationsmentioning

confidence: 99%

“…The skip connection has taken many kinds of advances during the evolution of UNet architecture. We have categorized the variation in the skip connection of the UNet architecture by considering nine new variations shown in Table 3, namely (S1) direct skip without any addition [44,46,51,76]; (S2) UNet++ with middle neurons are supported by the encoder layers [130,144,154]; (S3) UNet+++ (UNet 3+) skip connection with full scale [157]; (S4) addition of long short-term memory network (LSTM) [159], recurrent neural network (RNN) [109], generative adversial network (GAN), bidirectional long short-term memory network (BiLSTM), bidirectional recurrent neural network (BiRNN), bidirectional generative adversial network (BiGAN) [74,160]; (S5) skip connection between two cascaded UNet where skip of one is connected with skip of second and output of UNet first decoder goes in the skip of second UNet [81]; (S6) skip of first UNet into the skip of second UNet [89]; (S7) addition of the channel attention to improve the feature extraction along with atrous dense convolution block [65, 66, 71, 75, 82, 113, 119, 125, 126, 128, 129, 131-135, 139, 161]; (S8) squeeze-excitation (SE) network [125,133]; and (S9) Res skip [150]. The skip connection offers the ability to interface the powerful classification paradigms such as LSTM, RNN, and GAN.…”

Section: Skip Connection and Its Variationsmentioning

confidence: 99%

See 2 more Smart Citations

UNet Deep Learning Architecture for Segmentation of Vascular and Non-Vascular Images: A Microscopic Look at UNet Components Buffered With Pruning, Explainable Artificial Intelligence, and Bias

Suri¹,

Bhagawati

Agarwal³

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Background & Motivation: Biomedical image segmentation (BIS) task is challenging due to the variations in organ types, position, shape, size, scale, orientation, and image contrast. Conventional methods lack accurate and automated designs. Artificial intelligence (AI)-based UNet has recently dominated BIS. This is the first review of its kind that microscopically addressed UNet types by complexity, stratification of UNet by its components, addressing UNet in vascular vs. non-vascular framework, the key to segmentation challenge vs. UNet-based architecture, and finally interfacing the three facets of AI, the pruning, the explainable AI (XAI), and the AI-bias. Methods: PRISMA was used to select 267 UNet-based studies. Five classes were identified and labeled as conventional UNet, superior UNet, attention-channel UNet, hybrid UNet, and ensemble UNet. We discovered 81 variations of UNet by considering six kinds of components, namely encoder, decoder, skip connection, bridge network, loss function, and their combination. Vascular vs. non-vascular UNet architecture was compared. AP(ai)Bias 2.0-UNet was identified in these UNet classes based on (i) attributes of UNet architecture and its performance, (ii) explainable AI (XAI), and, (iii) pruning (compression). Five bias methods such as (i) ranking, (ii) radial, (iii) regional area, (iv) PROBAST, and (v) ROBINS-I were applied and compared using a Venn diagram. Findings and Conclusions: Vascular and non-vascular UNet systems dominated with sUNet classes with attention. The majority of the studies suffered from low interest in XAI and pruning strategies. None of the UNet models qualified to be bias-free. There is a need to move from paper-to-practice paradigms for clinical evaluation and settings. INDEX TERMS Image segmentation; vascular; non-vascular; UNet classes; UNet variations; UNetcomponents; explainable AI; pruning; bias.

show abstract

Section: B Five Types Of Unet and Their Attributesmentioning

confidence: 99%

Section: A Encoder Variationsmentioning

confidence: 99%

Section: Skip Connection and Its Variationsmentioning

confidence: 99%

See 1 more Smart Citation

UNet Deep Learning Architecture for Segmentation of Vascular and Non-Vascular Images: A Microscopic Look at UNet Components Buffered With Pruning, Explainable Artificial Intelligence, and Bias

Suri¹,

Bhagawati

Agarwal³

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Then, the decoder combines the bilinear (upsampling) and convolution layers to predict a binary image (the pixel value of the tongue region is 1, and the value of the other pixels is 0)

in size. There are some pipelines to transmit features and superimpose them on subsequent layers to enhance the information and resolution of the neural networks between the encoder and decoder [ 40 ].…”

Section: Tongue Processing Frameworkmentioning

confidence: 99%

An Intelligent Tongue Diagnosis System via Deep Learning on the Android Platform

Yang

Zhao

et al. 2022

Diagnostics

View full text Add to dashboard Cite

To quickly and accurately identify the pathological features of the tongue, we developed an intelligent tongue diagnosis system that uses deep learning on a mobile terminal. We also propose an efficient and accurate tongue image processing algorithm framework to infer the category of the tongue. First, a software system integrating registration, login, account management, tongue image recognition, and doctor–patient dialogue was developed based on the Android platform. Then, the deep learning models, based on the official benchmark models, were trained by using the tongue image datasets. The tongue diagnosis algorithm framework includes the YOLOv5s6, U-Net, and MobileNetV3 networks, which are employed for tongue recognition, tongue region segmentation, and tongue feature classification (tooth marks, spots, and fissures), respectively. The experimental results demonstrate that the performance of the tongue diagnosis model was satisfying, and the accuracy of the final classification of tooth marks, spots, and fissures was 93.33%, 89.60%, and 97.67%, respectively. The construction of this system has a certain reference value for the objectification and intelligence of tongue diagnosis.

show abstract

“…With the CNN method, not only identifying objects but also segmenting images into classified objects can be performed [ 8 ]. Performing segmentation according to the estimated histology with the CNN method has recently been investigated for colonic endoscopy [ 11 , 12 ]. For gastric endoscopy, we have not seen any systematic study of performing segmentation with artificial intelligence.…”

Section: Introductionmentioning

confidence: 99%

Objective Methods of 5-Aminolevulinic Acid-Based Endoscopic Photodynamic Diagnosis Using Artificial Intelligence for Identification of Gastric Tumors

Yamashita

Kurumi

Fujii

et al. 2022

JCM

View full text Add to dashboard Cite

Positive diagnoses of gastric tumors from photodynamic diagnosis (PDD) images after the administration of 5-aminolevulinic acid are subjectively identified by expert endoscopists. Objective methods of tumor identification are needed to reduce potential misidentifications. We developed two methods to identify gastric tumors from PDD images. Method one was applied to segmented regions in the PDD endoscopic image to determine the region in LAB color space to be attributed to tumors using a multi-layer neural network. Method two aimed to diagnose tumors and determine regions in the PDD endoscopic image attributed to tumors using the convoluted neural network method. The efficiencies of diagnosing tumors were 77.8% (7/9) and 93.3% (14/15) for method one and method two, respectively. The efficiencies of determining tumor region defined as the ratio of the area were 35.7% (0.0–78.0) and 48.5% (3.0–89.1) for method one and method two, respectively. False-positive rates defined as the ratio of the area were 0.3% (0.0–2.0) and 3.8% (0.0–17.4) for method one and method two, respectively. Objective methods of determining tumor region in 5-aminolevulinic acid-based endoscopic PDD were developed by identifying regions in LAB color space attributed to tumors or by applying a method of convoluted neural network.

show abstract

Use of U-Net Convolutional Neural Networks for Automated Segmentation of Fecal Material for Objective Evaluation of Bowel Preparation Quality in Colonoscopy

Cited by 10 publications

References 48 publications

UNet Deep Learning Architecture for Segmentation of Vascular and Non-Vascular Images: A Microscopic Look at UNet Components Buffered With Pruning, Explainable Artificial Intelligence, and Bias

UNet Deep Learning Architecture for Segmentation of Vascular and Non-Vascular Images: A Microscopic Look at UNet Components Buffered With Pruning, Explainable Artificial Intelligence, and Bias

An Intelligent Tongue Diagnosis System via Deep Learning on the Android Platform

Objective Methods of 5-Aminolevulinic Acid-Based Endoscopic Photodynamic Diagnosis Using Artificial Intelligence for Identification of Gastric Tumors

Contact Info

Product

Resources

About