Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles

Rajaraman, Sivaramakrishnan; Kim, In-Cheol; Antani, Sameer

doi:10.7717/peerj.8693

Cited by 31 publications

(16 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…CXR modality-specific pretraining: Previous studies reveal that compared to using ImageNet weights, CXR modality-specific model pretraining results in learning meaningful modality-specific features that can be transferred to improve performance in a relevant classification task [ 40 , 46 ]. We performed CXR modality-specific pretraining using a selection of various publicly available CXR data collections to introduce sufficient diversity into the training process in terms of acquisition methods, patient population, and other demographics, to help the models broadly learn significant features from CXRs showing normal lungs and other pulmonary abnormalities.…”

Section: Discussionmentioning

confidence: 99%

“…The modality-specific knowledge would be relevant to be transferred to the CXR classification task, as compared to using the ImageNet weights from the natural image processing domain. A previous study [ 40 ] shows the benefits of using CXR modality-specific models retraining toward improving classification and localization performance and model generalization. During CXR modality-specific pretraining, the data are split at the patient level into 90% for training and 10% for testing.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Chest X-ray Bone Suppression for Improving Classification of Tuberculosis-Consistent Findings

2021

Self Cite

View full text Add to dashboard Cite

Chest X-rays (CXRs) are the most commonly performed diagnostic examination to detect cardiopulmonary abnormalities. However, the presence of bony structures such as ribs and clavicles can obscure subtle abnormalities, resulting in diagnostic errors. This study aims to build a deep learning (DL)-based bone suppression model that identifies and removes these occluding bony structures in frontal CXRs to assist in reducing errors in radiological interpretation, including DL workflows, related to detecting manifestations consistent with tuberculosis (TB). Several bone suppression models with various deep architectures are trained and optimized using the proposed combined loss function and their performances are evaluated in a cross-institutional test setting using several metrics such as mean absolute error (MAE), peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and multiscale structural similarity measure (MS–SSIM). The best-performing model (ResNet–BS) (PSNR = 34.0678; MS–SSIM = 0.9828) is used to suppress bones in the publicly available Shenzhen and Montgomery TB CXR collections. A VGG-16 model is pretrained on a large collection of publicly available CXRs. The CXR-pretrained model is then fine-tuned individually on the non-bone-suppressed and bone-suppressed CXRs of Shenzhen and Montgomery TB CXR collections to classify them as showing normal lungs or TB manifestations. The performances of these models are compared using several performance metrics such as accuracy, the area under the curve (AUC), sensitivity, specificity, precision, F-score, and Matthews correlation coefficient (MCC), analyzed for statistical significance, and their predictions are qualitatively interpreted through class-selective relevance maps (CRMs). It is observed that the models trained on bone-suppressed CXRs (Shenzhen: AUC = 0.9535 ± 0.0186; Montgomery: AUC = 0.9635 ± 0.0106) significantly outperformed (p < 0.05) the models trained on the non-bone-suppressed CXRs (Shenzhen: AUC = 0.8991 ± 0.0268; Montgomery: AUC = 0.8567 ± 0.0870).. Models trained on bone-suppressed CXRs improved detection of TB-consistent findings and resulted in compact clustering of the data points in the feature space signifying that bone suppression improved the model sensitivity toward TB classification.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Chest X-ray Bone Suppression for Improving Classification of Tuberculosis-Consistent Findings

2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…Similarly another study revealed that deep learning algorithms can help in the identification of lung opacities, enlargement of heart and pleural effusion on chest X-rays [16]. Due to the better automated feature extraction capability, the convolutional neural networks which have been trained on natural images turned out to be very successful in the classification of images [17]. The performance of convolutional neural network related to the imaging modalities with less human intervention has been found to be better than other conventional methods of machine learning [18].…”

Section: Discussionmentioning

confidence: 99%

The diagnostic evaluation of Convolutional Neural Network (CNN) for the assessment of chest X-ray of patients infected with COVID-19

Suk¹,

Ssk

Syed

et al. 2020

Preprint

View full text Add to dashboard Cite

Introduction : The main target of COVID-19 is the lungs where it may cause pneumonia in severely ill patients. Chest X- ray is an important diagnostic test to assess the lung for the damaging effects of COVID-19. Many other microbial pathogens can also cause damage to lungs leading to pneumonia but there are certain radiological features which can favor the diagnosis of pneumonia caused by COVID-19. With the rising number of cases of COVID-19, it would be imperative to develop computer programs which may assist the health professionals in the prevailing scenario. Materials & Methods : A total of two hundred and seventy eight (278) images of chest X-rays have been assessed by applying ResNet-50 convolutional neural network architectures in the present study. The digital images were acquired from the public repositories provided by University of Montreal and National Institutes of Health. These digital images of Chest X-rays were divided into three groups labeled as normal, pneumonia and COVID- 19. The third group contains digital images of chest X-rays of patients diagnosed with COVID-19 infection while the second group contains images of lung with pneumonia caused by other pathogens. Results : The radiological images included in the data set are 89 images of lungs with COVID-19 infection, 93 images of lungs without any radiological abnormality and 96 images of patient with pneumonia caused by other pathogens. In this data set, 80% of the images were employed for training, and 20% for testing. A pre-trained (on ImageNet data set) ResNet-50 architecture was used to diagnose the cases of COVID-19 infections on lung Xray images. The analysis of the data revealed that computer vision based program achieved diagnostic accuracy of 98.18 %, and F1-score of 98.19. Conclusion : The performance of convolutional neural network regarding the differentiation of pulmonary changes caused by COVID-19 from the other type of pneumonias on digital images of the chest X- rays is excellent and it may be an extremely useful adjunct tool for the health professionals. Keywords: COVID-19, Convolutional Neural Network, ResNet-50

show abstract

“…Recent advances reported in the literature use conventional hand-crafted feature descriptors/classifiers and DL models for classifying CXRs as showing normal lungs or pulmonary TB manifestations [ 6 , 7 ]. These studies reveal that the DL models outperform conventional methods toward medical image classification tasks, particularly CXR analysis [ 8 ]. The authors [ 7 ] used the publicly available TB datasets, including the Shenzhen TB CXR, the Montgomery TB CXR [ 6 ], the Belarus TB CXR [ 9 ], and another private TB dataset from the Thomas Jefferson University Hospital, Philadelphia.…”

Section: Introductionmentioning

confidence: 99%

Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations

et al. 2021

Self Cite

View full text Add to dashboard Cite

Deep learning (DL) has drawn tremendous attention for object localization and recognition in both natural and medical images. U-Net segmentation models have demonstrated superior performance compared to conventional hand-crafted feature-based methods. Medical image modality-specific DL models are better at transferring domain knowledge to a relevant target task than those pretrained on stock photography images. This character helps improve model adaptation, generalization, and class-specific region of interest (ROI) localization. In this study, we train chest X-ray (CXR) modality-specific U-Nets and other state-of-the-art U-Net models for semantic segmentation of tuberculosis (TB)-consistent findings. Automated segmentation of such manifestations could help radiologists reduce errors and supplement decision-making while improving patient care and productivity. Our approach uses the publicly available TBX11K CXR dataset with weak TB annotations, typically provided as bounding boxes, to train a set of U-Net models. Next, we improve the results by augmenting the training data with weak localization, postprocessed into an ROI mask, from a DL classifier trained to classify CXRs as showing normal lungs or suspected TB manifestations. Test data are individually derived from the TBX11K CXR training distribution and other cross-institutional collections, including the Shenzhen TB and Montgomery TB CXR datasets. We observe that our augmented training strategy helped the CXR modality-specific U-Net models achieve superior performance with test data derived from the TBX11K CXR training distribution and cross-institutional collections (p < 0.05). We believe that this is the first study to i) use CXR modality-specific U-Nets for semantic segmentation of TB-consistent ROIs and ii) evaluate the segmentation performance while augmenting the training data with weak TB-consistent localizations.

show abstract

Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles

Cited by 31 publications

References 34 publications

Chest X-ray Bone Suppression for Improving Classification of Tuberculosis-Consistent Findings

Chest X-ray Bone Suppression for Improving Classification of Tuberculosis-Consistent Findings

The diagnostic evaluation of Convolutional Neural Network (CNN) for the assessment of chest X-ray of patients infected with COVID-19

Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations

Contact Info

Product

Resources

About