Deep Learning for Diagnosis and Segmentation of Pneumothorax: The Results on the Kaggle Competition and Validation Against Radiologists

Tolkachev, Alexey; Sirazitdinov, Ilyas; Kholiavchenko, Maksym; Mustafaev, Tamerlan; Ibragimov, Bulat

doi:10.1109/jbhi.2020.3023476

Cited by 42 publications

(22 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, several studies have been published regarding the prediction of pneumothorax location on chest X-ray images using deep learning 9 – 11 . One previous study performed pneumothorax segmentation using a framework combining UNet CNNs with various backbone networks 9 . The framework's performance, as measured by the Dice factor in the study, was 0.8574.…”

Section: Discussionmentioning

confidence: 99%

“…A number of studies have been conducted regarding the use of deep learning for pneumothorax classification with chest X-rays 4 – 8 . However, existing studies using deep- learning methods to detect the location of pneumothorax in chest X-rays are insufficient 9 – 11 , where these studies have used the convolution neural networks 12 (CNNs).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Detection of the location of pneumothorax in chest X-rays using small artificial neural networks and a simple training process

Cho

Kim

Lim

et al. 2021

Sci Rep

View full text Add to dashboard Cite

The purpose of this study was to evaluate the diagnostic performance achieved by using fully-connected small artificial neural networks (ANNs) and a simple training process, the Kim-Monte Carlo algorithm, to detect the location of pneumothorax in chest X-rays. A total of 1,000 chest X-ray images with pneumothorax were taken randomly from NIH (the National Institutes of Health) public image database and used as the training and test sets. Each X-ray image with pneumothorax was divided into 49 boxes for pneumothorax localization. For each of the boxes in the chest X-ray images contained in the test set, the area under the receiver operating characteristic (ROC) curve (AUC) was 0.882, and the sensitivity and specificity were 80.6% and 83.0%, respectively. In addition, a common currently used deep-learning method for image recognition, the convolution neural network (CNN), was also applied to the same dataset for comparison purposes. The performance of the fully-connected small ANN was better than that of the CNN. Regarding the diagnostic performances of the CNN with different activation functions, the CNN with a sigmoid activation function for fully-connected hidden nodes was better than the CNN with the rectified linear unit (RELU) activation function. This study showed that our approach can accurately detect the location of pneumothorax in chest X-rays, significantly reduce the time delay incurred when diagnosing urgent diseases such as pneumothorax, and increase the effectiveness of clinical practice and patient care.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Detection of the location of pneumothorax in chest X-rays using small artificial neural networks and a simple training process

Cho

Kim

Lim

et al. 2021

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Misclassification was still possible even after reviewing process. Secondly, the datasets for training, validation, and testing were relatively small compared to other studies [6,8,15,16,20]. Expansion of the datasets from our institution or using public datasets should be considered.…”

Section: Discussionmentioning

confidence: 99%

Automated Radiology Alert System for Pneumothorax Detection on Chest Radiographs Improves Efficiency and Diagnostic Performance

et al. 2021

View full text Add to dashboard Cite

We aimed to set up an Automated Radiology Alert System (ARAS) for the detection of pneumothorax in chest radiographs by a deep learning model, and to compare its efficiency and diagnostic performance with the existing Manual Radiology Alert System (MRAS) at the tertiary medical center. This study retrospectively collected 1235 chest radiographs with pneumothorax labeling from 2013 to 2019, and 337 chest radiographs with negative findings in 2019 were separated into training and validation datasets for the deep learning model of ARAS. The efficiency before and after using the model was compared in terms of alert time and report time. During parallel running of the two systems from September to October 2020, chest radiographs prospectively acquired in the emergency department with age more than 6 years served as the testing dataset for comparison of diagnostic performance. The efficiency was improved after using the model, with mean alert time improving from 8.45 min to 0.69 min and the mean report time from 2.81 days to 1.59 days. The comparison of the diagnostic performance of both systems using 3739 chest radiographs acquired during parallel running showed that the ARAS was better than the MRAS as assessed in terms of sensitivity (recall), area under receiver operating characteristic curve, and F1 score (0.837 vs. 0.256, 0.914 vs. 0.628, and 0.754 vs. 0.407, respectively), but worse in terms of positive predictive value (PPV) (precision) (0.686 vs. 1.000). This study had successfully designed a deep learning model for pneumothorax detection on chest radiographs and set up an ARAS with improved efficiency and overall diagnostic performance.

show abstract

“…Tolkachev [110] proposed a segmentation model for identifying the area of lung affected by pneumothorax using SIIM Pneumothorax dataset containing 12,047 CXRs. A U-Net model was used in which various CNNs including SE-ResNext50, ResNet34, DenseNet121 and SE-ResNext101 were used to replace the traditional encoder.…”

Section: • Localizationmentioning

confidence: 99%

“…On the other hand, DL algorithms do the job of feature extraction by itself. Despite the several limitations of DL technique such as requirement of huge amount of training data, computational complexity, extensive resource consumption and optimizing the hyperparameters, mostly researchers have adapted DL techniques because of their outstanding performance [22], [43], [61], [73], [110], [116]. • The performance of deep learning models can further be improved by ensemble modelling in which different models are trained on the training set and final results are combined via voting or averaging methods.…”

Section: A Comparative Analysismentioning

confidence: 99%

Automatic Diagnosis of Pneumothorax From Chest Radiographs: A Systematic Literature Review

Iqbal

Shaukat²,

Akram

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Among various medical imaging tools, chest radiographs are the most important and widely used diagnostic tool for the detection of thoracic pathologies. Remarkable research is being carried out to propose robust automatic diagnostic tools for the detection of pathologies from the chest radiographs. Artificial Intelligence techniques have been found to give promising results in automating the field of medicine. Lot of research has been done for automatic detection of pneumothorax from chest radiographs while proposing several frameworks based on artificial intelligence techniques. Undoubtedly, several models are available for automatic diagnosis of pneumothorax, however a summarized review of the existing literature is still missing. This study summarizes the existing literature for pneumothorax detection from chest x-rays along with describing the available chest radiographs datasets. It will help the researchers to select the optimal and most effective model with respect to the real-time scenarios. The comparative analysis of the literature is provided in terms of goodness, usability and quality along with highlighting the research gaps for further investigation. From the literature, it is evident that pneumothorax is more common in men as compared to women. Additionally, the proposed models have achieved incredible results for pneumothorax detection on selected datasets, however, the effectiveness of proposed models in real-time cases cannot be claimed, as none of the models have been implemented clinically yet. This issue can be solved by external validation of the models. Furthermore, the class-imbalance problem in most of the medical image dataset has been solved by algorithm-level-techniques. Moreover, there is need to put more effort in combined detection and localization of pneumothorax as mostly research is limited to either classification or localization of pathology. So far, best results have been achieved by deep-learning based models with Area-underreceiver-operating-characteristic-curve (AUC) of 88.87% for classification, and Dice-similarity-coefficient (DSC) of 88.21% for localization of pneumothorax. Thus, the outstanding abilities of deep learning techniques can be deployed for developing robust models for pneumothorax detection.

show abstract

Deep Learning for Diagnosis and Segmentation of Pneumothorax: The Results on the Kaggle Competition and Validation Against Radiologists

Cited by 42 publications

References 26 publications

Detection of the location of pneumothorax in chest X-rays using small artificial neural networks and a simple training process

Detection of the location of pneumothorax in chest X-rays using small artificial neural networks and a simple training process

Automated Radiology Alert System for Pneumothorax Detection on Chest Radiographs Improves Efficiency and Diagnostic Performance

Automatic Diagnosis of Pneumothorax From Chest Radiographs: A Systematic Literature Review

Contact Info

Product

Resources

About