Joint Learning for Pneumonia Classification and Segmentation on Medical Images

2020 ◽  
pp. 2157003
Author(s):  
Shaobo Liu ◽  
Xin Zhong ◽  
Frank Y. Shih
Keyword(s):  
Medical Images ◽  
Saliency Map ◽  
Classification Model ◽  
Automatic Identification ◽  
Test Accuracy ◽  
Joint Learning ◽  
Joint Task ◽  
Chest X Ray ◽  
Task Architecture ◽  
Weakly Supervised

Chest X-ray images are notoriously difficult to analyze due to the noisy nature. Automatic identification of pneumonia on medical images has attracted intensive study recently. In this paper, a novel joint-task architecture that can learn pneumonia classification and segmentation simultaneously is presented. Two modules, including an image preprocessing module and an attention module, are developed to improve both the classification and segmentation accuracies. Results from the experiments performed on the massive dataset of the Radiology Society of North America have confirmed its superiority over the other existing methods. The classification test accuracy is improved from 0.89 to 0.95, and the segmentation model achieves an improved mean precision result of 0.58–0.78. Finally, two weakly supervised learning methods, class-saliency map and Grad-CAM, are used to highlight the corresponding pixels or areas which have significant influence on the classification model, such that the refined segmentation can focus on the correct areas with high confidence.

scholarly journals Optimal Randomness for Stochastic Configuration Network (SCN) with Heavy-Tailed Distributions

Entropy ◽  
2020 ◽  
Vol 23 (1) ◽  
pp. 56
Author(s):  
Haoyu Niu ◽  
Jiamin Wei ◽  
YangQuan Chen
Keyword(s):  
Research Work ◽  
Cauchy Distribution ◽  
Classification Model ◽  
Test Accuracy ◽  
Functional Link ◽  
Heavy Tailed Distributions ◽  
Heavy Tailed ◽  
Improved Performance ◽  
Hidden Layer ◽  
Hidden Nodes

Stochastic Configuration Network (SCN) has a powerful capability for regression and classification analysis. Traditionally, it is quite challenging to correctly determine an appropriate architecture for a neural network so that the trained model can achieve excellent performance for both learning and generalization. Compared with the known randomized learning algorithms for single hidden layer feed-forward neural networks, such as Randomized Radial Basis Function (RBF) Networks and Random Vector Functional-link (RVFL), the SCN randomly assigns the input weights and biases of the hidden nodes in a supervisory mechanism. Since the parameters in the hidden layers are randomly generated in uniform distribution, hypothetically, there is optimal randomness. Heavy-tailed distribution has shown optimal randomness in an unknown environment for finding some targets. Therefore, in this research, the authors used heavy-tailed distributions to randomly initialize weights and biases to see if the new SCN models can achieve better performance than the original SCN. Heavy-tailed distributions, such as Lévy distribution, Cauchy distribution, and Weibull distribution, have been used. Since some mixed distributions show heavy-tailed properties, the mixed Gaussian and Laplace distributions were also studied in this research work. Experimental results showed improved performance for SCN with heavy-tailed distributions. For the regression model, SCN-Lévy, SCN-Mixture, SCN-Cauchy, and SCN-Weibull used less hidden nodes to achieve similar performance with SCN. For the classification model, SCN-Mixture, SCN-Lévy, and SCN-Cauchy have higher test accuracy of 91.5%, 91.7% and 92.4%, respectively. Both are higher than the test accuracy of the original SCN.

scholarly journals Learning Hierarchical Attention for Weakly-supervised Chest X-Ray Abnormality Localization and Diagnosis

2020 ◽  
pp. 1-1
Author(s):  
Xi Ouyang ◽  
Srikrishna Karanam ◽  
Ziyan Wu ◽  
Terrence Chen ◽  
Jiayu Huo ◽  
...  
Keyword(s):  
X Ray ◽  
Chest X Ray ◽  
Weakly Supervised

Optimal Scale of Hierarchical Image Segmentation with Scribbles Guidance for Weakly Supervised Semantic Segmentation

2021 ◽  
pp. 2154026
Author(s):  
Zaid Al-Huda ◽  
Donghai Zhai ◽  
Yan Yang ◽  
Riyadh Nazar Ali Algburi
Keyword(s):  
Image Segmentation ◽  
Graphical Model ◽  
Semantic Segmentation ◽  
Saliency Map ◽  
Training Data ◽  
Deep Convolutional Neural Networks ◽  
High Quality ◽  
Optimal Scale ◽  
Supervised Segmentation ◽  
Weakly Supervised

Deep convolutional neural networks (DCNNs) trained on the pixel-level annotated images have achieved improvements in semantic segmentation. Due to the high cost of labeling training data, their applications may have great limitation. However, weakly supervised segmentation approaches can significantly reduce human labeling efforts. In this paper, we introduce a new framework to generate high-quality initial pixel-level annotations. By using a hierarchical image segmentation algorithm to predict the boundary map, we select the optimal scale of high-quality hierarchies. In the initialization step, scribble annotations and the saliency map are combined to construct a graphic model over the optimal scale segmentation. By solving the minimal cut problem, it can spread information from scribbles to unmarked regions. In the training process, the segmentation network is trained by using the initial pixel-level annotations. To iteratively optimize the segmentation, we use a graphical model to refine segmentation masks and retrain the segmentation network to get more precise pixel-level annotations. The experimental results on Pascal VOC 2012 dataset demonstrate that the proposed framework outperforms most of weakly supervised semantic segmentation methods and achieves the state-of-the-art performance, which is [Formula: see text] mIoU.

scholarly journals Improving Joint Learning of Chest X-Ray and Radiology Report by Word Region Alignment

2021 ◽  
pp. 110-119
Author(s):  
Zhanghexuan Ji ◽  
Mohammad Abuzar Shaikh ◽  
Dana Moukheiber ◽  
Sargur N Srihari ◽  
Yifan Peng ◽  
...  
Keyword(s):  
Radiology Report ◽  
Joint Learning ◽  
X Ray ◽  
Chest X Ray

scholarly journals COVID-19 prediction using chest X-ray medical images and Deep Learning

2021 ◽  
Author(s):  
Nuria Pereira Espasandín ◽  
David Maseda Neira ◽  
Diana Marcela Noriega Cobo ◽  
Iago Iglesias Corrás ◽  
Alejandro Pazos ◽  
...  
Keyword(s):  
Deep Learning ◽  
Medical Images ◽  
X Ray ◽  
Chest X Ray

scholarly journals ORFhunteR: an accurate approach for the automatic identification and annotation of open reading frames in human mRNA molecules

2021 ◽  
Author(s):  
Vasily V. Grinev ◽  
Mikalai M. Yatskou ◽  
Victor V. Skakun ◽  
Maryna K. Chepeleva ◽  
Petr V. Nazarov
Keyword(s):  
Single Molecule ◽  
Web Application ◽  
R Package ◽  
Nucleotide Sequences ◽  
Open Reading Frames ◽  
Classification Model ◽  
Automatic Identification ◽  
Large Set ◽  
Link Type ◽  
Reading Frames

AbstractMotivationModern methods of whole transcriptome sequencing accurately recover nucleotide sequences of RNA molecules present in cells and allow for determining their quantitative abundances. The coding potential of such molecules can be estimated using open reading frames (ORF) finding algorithms, implemented in a number of software packages. However, these algorithms show somewhat limited accuracy, are intended for single-molecule analysis and do not allow selecting proper ORFs in the case of long mRNAs containing multiple ORF candidates.ResultsWe developed a computational approach, corresponding machine learning model and a package, dedicated to automatic identification of the ORFs in large sets of human mRNA molecules. It is based on vectorization of nucleotide sequences into features, followed by classification using a random forest. The predictive model was validated on sets of human mRNA molecules from the NCBI RefSeq and Ensembl databases and demonstrated almost 95% accuracy in detecting true ORFs. The developed methods and pre-trained classification model were implemented in a powerful ORFhunteR computational tool that performs an automatic identification of true ORFs among large set of human mRNA molecules.Availability and implementationThe developed open-source R package ORFhunteR is available for the community at GitHub repository (https://github.com/rfctbio-bsu/ORFhunteR), from Bioconductor (https://bioconductor.org/packages/devel/bioc/html/ORFhunteR.html) and as a web application (http://orfhunter.bsu.by).

scholarly journals MFDNN: Multi-channel feature deep neural network algorithm to identify Covid19 chest X-ray images

2021 ◽  
Author(s):  
Liangrui Pan ◽  
boya ji ◽  
Xiaoqi wang ◽  
shaoliang peng
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Deep Neural Network ◽  
Test Accuracy ◽  
Learning Networks ◽  
X Ray ◽  
Network Algorithm ◽  
Life Saving ◽  
Chest X Ray ◽  
Neural Network Algorithm

The use of chest X-ray images (CXI) to detect Severe Acute Respiratory Syndrome Coronavirus 2 (SARS CoV-2) caused by Coronavirus Disease 2019 (COVID-19) is life-saving important for both patients and doctors. This research proposed a multi-channel feature deep neural network algorithm to screen people infected with COVID-19. The algorithm integrates data oversampling technology and a multi-channel feature deep neural network model to carry out the training process in an end-to-end manner. In the experiment, we used a publicly available CXI database with 10,192 Normal, 6012 Lung Opacity (Non-COVID lung infection), and 1345 Viral Pneumonia images. Compared with traditional deep learning models (Densenet201, ResNet50, VGG19, GoogLeNet), the MFDNN model obtains an average test accuracy of 93.19% in all data. Furthermore, in each type of screening, the precision, recall, and F1 Score of the MFDNN model are also better than traditional deep learning networks. Secondly, compared with the latest CoroDet model, the MFDNN algorithm is 1.91% higher than the CoroDet model in the experiment of detecting the four categories of COVID19 infected persons. Finally, our experimental code will be placed at https://github.com/panliangrui/covid19.

scholarly journals Weakly Supervised Contrastive Learning for Chest X-Ray Report Generation

2021 ◽  
Author(s):  
An Yan ◽  
Zexue He ◽  
Xing Lu ◽  
Jiang Du ◽  
Eric Chang ◽  
...  
Keyword(s):  
Report Generation ◽  
X Ray ◽  
Chest X Ray ◽  
Weakly Supervised

Alternate methods for anomaly detection in high-energy physics via semi-supervised learning

2020 ◽  
Vol 35 (23) ◽  
pp. 2050131
Author(s):  
Mohd Adli Md Ali ◽  
Nu’man Badrud’din ◽  
Hafidzul Abdullah ◽  
Faiz Kemi
Keyword(s):  
Anomaly Detection ◽  
Supervised Learning ◽  
High Energy Physics ◽  
High Energy ◽  
Classification Model ◽  
Training Dataset ◽  
Weakly Supervised Learning ◽  
Physics Community ◽  
Weakly Supervised ◽  
Energy Physics

Recently, the concept of weakly supervised learning has gained popularity in the high-energy physics community due to its ability to learn even with a noisy and impure dataset. This method is valuable in the quest to discover the elusive beyond Standard Model (BSM) particle. Nevertheless, the weakly supervised learning method still requires a learning sample that describes the features of the BSM particle truthfully to the classification model. Even with the various theoretical framework such as supersymmetry and the quantum black hole, creating a BSM sample is not a trivial task since the exact feature of the particle is unknown. Due to these difficulties, we propose an alternative classifier type called the one-class classification (OCC). OCC algorithms require only background or noise samples in its training dataset, which is already abundant in the high-energy physics community. The algorithm will flag any sample that does not fit the background feature as an abnormality. In this paper, we introduce two new algorithms called EHRA and C-EHRA, which use machine learning regression and clustering to detect anomalies in samples. We tested the algorithms’ capability to create distinct anomalous patterns in the presence of BSM samples and also compare their classification output metrics to the Isolation Forest (ISF), a well-known anomaly detection algorithm. Five Monte Carlo supersymmetry datasets with the signal to noise ratio equal to 1, 0.1, 0.01, 0.001, and 0.0001 were used to test EHRA, C-EHRA and ISF algorithm. In our study, we found that the EHRA with an artificial neural network regression has the highest ROC-AUC score at 0.7882 for the balanced dataset, while the C-EHRA has the highest precision-sensitivity score for the majority of the imbalanced datasets. These findings highlight the potential use of the EHRA, C-EHRA, and other OCC algorithms in the quest to discover BSM particles.

Sign in / Sign up

Export Citation Format

Share Document