scholarly journals Automated Diagnosis of COVID-19 Using Deep Learning and Data Augmentation on Chest CT

2020 ◽  
Author(s):  
Runwen Hu ◽  
Guanqi Ruan ◽  
Shijun Xiang ◽  
Minghui Huang ◽  
Qiaoyi Liang ◽  
...  
Keyword(s):  
Deep Learning ◽  
Receiver Operating Characteristic ◽  
Operating Characteristic ◽  
Data Augmentation ◽  
Chest Ct ◽  
Training Set ◽  
Sensitivity Specificity ◽  
Noisy Labels ◽  
Testing Set

AbstractBackgroundCoronavirus disease 2019 (COVID-19) has surprised the world since the beginning of 2020, and the rapid growth of COVID-19 is beyond the capability of doctors and hospitals that could deal in many areas. The chest computed tomography (CT) could be served as an effective tool in detection of COVID-19. It is valuable to develop automatic detection of COVID-19.Materials and MethodsThe collected dataset consisted of 1042 chest CT images (including 521 COVID-19, 397 healthy, 76 bacterial pneumonia and 48 SARS) obtained by exhaustively searching available data on the Internet. Then, these data are divided into three sets, referred to training set, validation set and testing set. Sixteen data augmentation operations are designed to enrich the training set in deep learning training phase. Multiple experiments were conducted to analyze the performance of the model in the detection of COVID-19 both in case of no noisy labels and noisy labels. The performance was assessed by the area under the receiver operating characteristic (AUC), sensitivity, specificity and accuracy.ResultsThe data augmentation operations on the training set are effective for improvement of the model performance. The area under the receiver operating characteristic curve is 0.9689 with (95% CI: 0.9308, 1) in case of no noisy labels for the classification of COVID-19 from heathy subject, while the per-exam sensitivity, specificity and accuracy for detecting COVID-19 in the independent testing set are 90.52%, 91.58% and 91.21%, respectively. In the classification of COVID-19 from other hybrid cases, the average AUC of the proposed model is 0.9222 with (95%CI: 0.8418, 1) if there are no noisy labels. The model is also robust when part of the training samples is marked incorrectly. The average AUC is 92.23% in the case of noisy labels of 10% in the training set.ConclusionA deep learning model with insufficient samples can be developed by using data augmentation in assisting medical workers in making quick and correct diagnosis of COVID-19.

scholarly journals Acute Myocardial Infarction Detection Using Deep Learning-Enabled Electrocardiograms

2021 ◽  
Vol 8 ◽  
Author(s):  
Xiehui Chen ◽  
Wenqin Guo ◽  
Lingyue Zhao ◽  
Weichao Huang ◽  
Lili Wang ◽  
...  
Keyword(s):  
Myocardial Infarction ◽  
Acute Myocardial Infarction ◽  
Deep Learning ◽  
Model Performance ◽  
Area Under The Curve ◽  
Training Set ◽  
Residual Network ◽  
Validation Set ◽  
Sensitivity Specificity ◽  
Testing Set

Background: Acute myocardial infarction (AMI) is associated with a poor prognosis. Therefore, accurate diagnosis and early intervention of the culprit lesion are of extreme importance. Therefore, we developed a neural network algorithm in this study to automatically diagnose AMI from 12-lead electrocardiograms (ECGs).Methods: We used the open-source PTB-XL database as the training and validation sets, with a 7:3 sample size ratio. Twenty-One thousand, eight hundred thirty-seven clinical 12-lead ECGs from the PTB-XL dataset were available for training and validation (15,285 were used in the training set and 6,552 in the validation set). Additionally, we randomly selected 205 ECGs from a dataset built by Chapman University, CA, USA and Shaoxing People's Hospital, China, as the testing set. We used a residual network for training and validation. The model performance was experimentally verified in terms of area under the curve (AUC), precision, sensitivity, specificity, and F1 score.Results: The AUC of the training, validation, and testing sets were 0.964 [95% confidence interval (CI): 0.961–0.966], 0.944 (95% CI: 0.939–0.949), and 0.977 (95% CI: 0.961–0.991), respectively. The precision, sensitivity, specificity, and F1 score of the deep learning model for AMI diagnosis from ECGs were 0.827, 0.824, 0.950, and 0.825, respectively, in the training set, 0.789, 0.818, 0.913, and 0.803, respectively, in the validation set, and 0.830, 0.951, 0.951, and 0.886, respectively, in the testing set. The AUC for automatic AMI location diagnosis of LMI, IMI, ASMI, AMI, ALMI were 0.969 (95% CI: 0.959–0.979), 0.973 (95% CI: 0.962–0.978), 0.987 (95% CI: 0.963–0.989), 0.961 (95% CI: 0.956–0.989), and 0.996 (95% CI: 0.957–0.997), respectively.Conclusions: The residual network-based algorithm can effectively automatically diagnose AMI and MI location from 12-lead ECGs.

scholarly journals An Imbalanced Image Classification Method for the Cell Cycle Phase

Information ◽  
2021 ◽  
Vol 12 (6) ◽  
pp. 249
Author(s):  
Xin Jin ◽  
Yuanwen Zou ◽  
Zhongbing Huang
Keyword(s):  
Cell Cycle ◽  
Deep Learning ◽  
Image Classification ◽  
Classification Accuracy ◽  
Data Augmentation ◽  
Cycle Phase ◽  
Generative Adversarial Network ◽  
Adversarial Network ◽  
Cellular Life

The cell cycle is an important process in cellular life. In recent years, some image processing methods have been developed to determine the cell cycle stages of individual cells. However, in most of these methods, cells have to be segmented, and their features need to be extracted. During feature extraction, some important information may be lost, resulting in lower classification accuracy. Thus, we used a deep learning method to retain all cell features. In order to solve the problems surrounding insufficient numbers of original images and the imbalanced distribution of original images, we used the Wasserstein generative adversarial network-gradient penalty (WGAN-GP) for data augmentation. At the same time, a residual network (ResNet) was used for image classification. ResNet is one of the most used deep learning classification networks. The classification accuracy of cell cycle images was achieved more effectively with our method, reaching 83.88%. Compared with an accuracy of 79.40% in previous experiments, our accuracy increased by 4.48%. Another dataset was used to verify the effect of our model and, compared with the accuracy from previous results, our accuracy increased by 12.52%. The results showed that our new cell cycle image classification system based on WGAN-GP and ResNet is useful for the classification of imbalanced images. Moreover, our method could potentially solve the low classification accuracy in biomedical images caused by insufficient numbers of original images and the imbalanced distribution of original images.

scholarly journals UAV Image Multi-Labeling with Data-Efficient Transformers

2021 ◽  
Vol 11 (9) ◽  
pp. 3974
Author(s):  
Laila Bashmal ◽  
Yakoub Bazi ◽  
Mohamad Mahmoud Al Rahhal ◽  
Haikel Alhichri ◽  
Naif Al Ajlan
Keyword(s):  
Data Augmentation ◽  
Feature Representation ◽  
Aerial Image ◽  
Remote Sensing Images ◽  
Training Set ◽  
Proposed Model ◽  
Class Labels ◽  
Using Data ◽  
Uav Image

In this paper, we present an approach for the multi-label classification of remote sensing images based on data-efficient transformers. During the training phase, we generated a second view for each image from the training set using data augmentation. Then, both the image and its augmented version were reshaped into a sequence of flattened patches and then fed to the transformer encoder. The latter extracts a compact feature representation from each image with the help of a self-attention mechanism, which can handle the global dependencies between different regions of the high-resolution aerial image. On the top of the encoder, we mounted two classifiers, a token and a distiller classifier. During training, we minimized a global loss consisting of two terms, each corresponding to one of the two classifiers. In the test phase, we considered the average of the two classifiers as the final class labels. Experiments on two datasets acquired over the cities of Trento and Civezzano with a ground resolution of two-centimeter demonstrated the effectiveness of the proposed model.

scholarly journals Diagnostic performances of manual and automated reticulocyte parameters in anaemic cats

2017 ◽  
Vol 20 (2) ◽  
pp. 122-127 ◽  
Author(s):  
Saverio Paltrinieri ◽  
Marco Fossati ◽  
Valentina Menaballi
Keyword(s):  
Receiver Operating Characteristic ◽  
Sensitivity And Specificity ◽  
Operating Characteristic ◽  
Positive Likelihood Ratio ◽  
Roc Curves ◽  
Blood Samples ◽  
Production Index ◽  
Instrumental Measurement ◽  
Sensitivity Specificity ◽  
Reticulocyte Production

Objectives The objective of this study was to evaluate the diagnostic performances of manual and instrumental measurement of reticulocyte percentage (Ret%), reticulocyte number (Ret#) and reticulocyte production index (RPI) to differentiate regenerative anaemia (RA) from non-regenerative anaemia (NRA) in cats. Methods Data from 106 blood samples from anaemic cats with manual counts (n = 74; 68 NRA, six RA) or instrumental counts of reticulocytes (n = 32; 25 NRA, seven RA) collected between 1995 and 2013 were retrospectively analysed. Sensitivity, specificity and positive likelihood ratio (LR+) were calculated using either cut-offs reported in the literature or cut-offs determined from receiver operating characteristic (ROC) curves. Results All the reticulocyte parameters were significantly higher in cats with RA than in cats with NRA. All the ROC curves were significantly different ( P <0.001) from the line of no discrimination, without significant differences between the three parameters. Using the cut-offs published in literature, the Ret% (cut-off: 0.5%) was sensitive (100%) but not specific (<75%), the RPI (cut-off: 1.0) was specific (>92%) but not sensitive (<15%), and the Ret# (cut-off: 50 × 10³/µl) had a sensitivity and specificity >80% and the highest LR+ (manual count: 14; instrumental count: 6). For all the parameters, sensitivity and specificity approached 100% using the cut-offs determined by the ROC curves. These cut-offs were higher than those reported in the literature for Ret% (manual: 1.70%; instrumental: 3.06%), lower for RPI (manual: 0.39; instrumental: 0.59) and variably different, depending on the method (manual: 41 × 10³/µl; instrumental: 57 × 10³/µl), for Ret#. Using these cut-offs, the RPI had the highest LR+ (manual: 22.7; instrumental: 12.5). Conclusions and relevance This study indicated that all the reticulocyte parameters may confirm regeneration when the pretest probability is high, while when this probability is moderate, RA should be identified using the RPI providing that cut-offs <1.0 are used.

scholarly journals Bayesian Sensitivity-Specificity and ROC Analysis for Finding Key Drivers

2021 ◽  
Vol 19 (1) ◽  
pp. 2-15
Author(s):  
Stan Lipovetsky ◽  
Michael W. Conklin
Keyword(s):  
Receiver Operating Characteristic ◽  
Statistical Modeling ◽  
Roc Analysis ◽  
Operating Characteristic ◽  
Regression Modeling ◽  
Modeling And Analysis ◽  
Numerical Comparisons ◽  
Key Drivers ◽  
Sensitivity Specificity ◽  
Receiver Operating

Finding key drivers in regression modeling via Bayesian Sensitivity-Specificity and Receiver Operating Characteristic is suggested, and clearly interpretable results are obtained. Numerical comparisons with other techniques show that this methodology can be useful in practical statistical modeling and analysis helping to researchers and managers in making meaningful decisions.

scholarly journals Detection of Incidental Esophageal Cancers on Chest CT by Deep Learning

2021 ◽  
Vol 11 ◽  
Author(s):  
He Sui ◽  
Ruhang Ma ◽  
Lin Liu ◽  
Yaozong Gao ◽  
Wenhai Zhang ◽  
...  
Keyword(s):  
Esophageal Cancer ◽  
Deep Learning ◽  
Cancer Detection ◽  
Ct Images ◽  
Chest Ct ◽  
Validation Dataset ◽  
Esophageal Wall ◽  
Detection Model ◽  
Model Training ◽  
Sensitivity Specificity

ObjectiveTo develop a deep learning-based model using esophageal thickness to detect esophageal cancer from unenhanced chest CT images.MethodsWe retrospectively identified 141 patients with esophageal cancer and 273 patients negative for esophageal cancer (at the time of imaging) for model training. Unenhanced chest CT images were collected and used to build a convolutional neural network (CNN) model for diagnosing esophageal cancer. The CNN is a VB-Net segmentation network that segments the esophagus and automatically quantifies the thickness of the esophageal wall and detect positions of esophageal lesions. To validate this model, 52 false negatives and 48 normal cases were collected further as the second dataset. The average performance of three radiologists and that of the same radiologists aided by the model were compared.ResultsThe sensitivity and specificity of the esophageal cancer detection model were 88.8% and 90.9%, respectively, for the validation dataset set. Of the 52 missed esophageal cancer cases and the 48 normal cases, the sensitivity, specificity, and accuracy of the deep learning esophageal cancer detection model were 69%, 61%, and 65%, respectively. The independent results of the radiologists had a sensitivity of 25%, 31%, and 27%; specificity of 78%, 75%, and 75%; and accuracy of 53%, 54%, and 53%. With the aid of the model, the results of the radiologists were improved to a sensitivity of 77%, 81%, and 75%; specificity of 75%, 74%, and 74%; and accuracy of 76%, 77%, and 75%, respectively.ConclusionsDeep learning-based model can effectively detect esophageal cancer in unenhanced chest CT scans to improve the incidental detection of esophageal cancer.

scholarly journals Deep Learning Method for Quantitative Analysis of Covid -19 using CT Chest Images

2021 ◽  
Author(s):  
Indrajeet Kumar ◽  
Jyoti Rawat
Keyword(s):  
Deep Learning ◽  
Diagnostic Tests ◽  
Ct Images ◽  
Rapid Diagnosis ◽  
Chest Ct ◽  
Visual Features ◽  
Learning Module ◽  
Volumetric Ct ◽  
The Cost

Abstract The manual diagnostic tests performed in laboratories for pandemic disease such as COVID19 is time-consuming, requires skills and expertise of the performer to yield accurate results. Moreover, it is very cost ineffective as the cost of test kits is high and also requires well-equipped labs to conduct them. Thus, other means of diagnosing the patients with presence of SARS-COV2 (the virus responsible for COVID19) must be explored. A radiography method like chest CT images is one such means that can be utilized for diagnosis of COVID19. The radio-graphical changes observed in CT images of COVID19 patient helps in developing a deep learning-based method for extraction of graphical features which are then used for automated diagnosis of the disease ahead of laboratory-based testing. The proposed work suggests an Artificial Intelligence (AI) based technique for rapid diagnosis of COVID19 from given volumetric CT images of patient’s chest by extracting its visual features and then using these features in the deep learning module. The proposed convolutional neural network is deployed for classifying the infectious and non-infectious SARS-COV2 subjects. The proposed network utilizes 746 chests scanned CT images of which 349 images belong to COVID19 positive cases while remaining 397 belong negative cases of COVID19. The extensive experiment has been completed with the accuracy of 98.4 %, sensitivity of 98.5 %, the specificity of 98.3 %, the precision of 97.1 %, F1score of 97.8 %. The obtained result shows the outstanding performance for classification of infectious and non-infectious for COVID19 cases.

scholarly journals Systematic Review with Meta-Analysis: Fecal Calprotectin as a Surrogate Marker for Predicting Relapse in Adults with Ulcerative Colitis

2019 ◽  
Vol 2019 ◽  
pp. 1-10 ◽  
Author(s):  
Jiajia Li ◽  
Xiaojing Zhao ◽  
Xueting Li ◽  
Meijiao Lu ◽  
Hongjie Zhang
Keyword(s):  
Ulcerative Colitis ◽  
Receiver Operating Characteristic ◽  
Likelihood Ratio ◽  
Odds Ratio ◽  
Operating Characteristic ◽  
Diagnostic Odds Ratio ◽  
Fecal Calprotectin ◽  
Clinical Relapse ◽  
Sensitivity Specificity ◽  
Receiver Operating

The clinical course of ulcerative colitis (UC) is featured by remission and relapse, which remains unpredictable. Recent studies revealed that fecal calprotectin (FC) could predict clinical relapse for UC patients in remission, which has not yet been well accepted. To detect the predictive value of FC for clinical relapse in adult UC patients based on updated literature, we carried out a comprehensive electronic search of PubMed, Web of Science, Embase, and the Cochrane Library to identify all eligible studies. Diagnostic accuracy including pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and pooled area under the receiver operating characteristic (AUROC) was calculated using a random effects model. Heterogeneity across studies was assessed by the I2 metric. Sources of heterogeneity were detected using subgroup analysis. Metaregression was used to test potential factors correlated to DOR. Publication bias was assessed using Deek’s funnel plots. In our study, 14 articles enrolling a total of 1110 participants were finally included, and all articles underwent a quality assessment. Pooled sensitivity, specificity, PLR, and NLR with 95% confidence intervals (CIs) were 0.75 (95% CI: 0.70–0.79), 0.77 (95% CI: 0.74–0.80), 3.45 (95% CI: 2.31–5.14), and 0.37 (95% CI: 0.28–0.49) respectively. The area under the summary receiver operating characteristic (sROC) curve was 0.82, and the diagnostic odds ratio was 10.54 (95% CI: 6.16–18.02). Our study suggested that FC is useful in predicting clinical relapse for adult UC patients in remission as a simple and noninvasive marker.

Sign in / Sign up

Export Citation Format

Share Document