Cardiomegaly Detection on Chest Radiographs: Segmentation Versus Classification

Sogancioglu, Ecem; Murphy, Keelin; Çallı, Erdi; Scholten, Ernst Th.; Schalekamp, Steven; Ginneken, Bram van

doi:10.1109/access.2020.2995567

Cited by 37 publications

(33 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This indicates that most of these categorical disagreements are simply due to CTR estimation noise around the 0.5 threshold (borderline cases) rather than significant departures in estimation between model and manual method. These near-boundary error cases were forgivable and mentioned in another study [9].…”

Section: ) Error Analysismentioning

confidence: 66%

“…Similarly, two human readers have an MAE of 0.0123, or a mean difference of -0.0076 ± 0.0136, with 95% limits of agreement from -0.0192 to 0.0343. Two other studies [7], [9] have benchmarked the performance of algorithms against trained CXR readers with comparable performances to our own. However, our study is, to the best of our knowledge, the only study where testing was performed on an external validation dataset (i.e.…”

Section: B Model Performance Benchmarked Against Manual Methods 1) Comparison Of Ctr Mean Errormentioning

confidence: 77%

“…The authors also compared the automatically generated CTRs with those obtained from human annotators with 0.87-0.97 Pearson's correlations. Sogancioglu et al (2020) [9] compared segmentation and classification approaches to cardiomegaly detection on the re-annotated ChestX-ray14 dataset. In relation to our paper and similarly to [6], [7], two U-Net networks were trained to segment lungs and hearts.…”

Section: Deep Learning Approaches To Ctr Estimationmentioning

confidence: 99%

See 2 more Smart Citations

External Validation of Deep Learning Algorithms for Cardiothoracic Ratio Measurement

et al. 2021

View full text Add to dashboard Cite

Recent advances in machine learning have made it possible to create automated systems for medical image diagnosis. Cardiothoracic ratio (CTR) measurement, a common procedure for assessing cardiac abnormality in chest radiographs, has been investigated by several deep learning studies aiming to automate the process. However, of key consideration is whether automated CTR measurements by machine learning models can yield CTR values as accurately and consistently as trained human technicians on unseen data and thereby be considered trustworthy for clinical application. To assess this, we performed external validation of automated CTR algorithms on a dataset of 7,517 images, comparing four variants of U-Net architecture in heart and lung segmentations and CTR calculations: VGG-11, VGG-16, SegNet, and AlbuNet. We then benchmarked their performance against two human experts manually measuring CTR on the same images in a clinical setting such that we could equitably compare model-to-human variation against human-to-human variation. Our analysis shows that AlbuNet demonstrates human-level performance in CTR measurements, achieving MAPE of 2.38%, which is on par with human-to-human inter-rater variability (2.53%) when using the manual measurement method. The other three U-Net variants, particularly VGG-16, also performed similarly well. Additionally, we conducted an extreme outlier analysis on each model architecture, assessing the percentage of samples with higher measurement errors than the maximum error from the manual method. AlbuNet outperformed other architectures with only 0.35% occurrence of extreme outliers, while the other three U-Net variants ranged from 0.64% to 1.06% occurrence. Overall, the deep-learning-based algorithm was demonstrated to be as reliable as the manual method and shows strong potential for assisting radiologists in the CTR measurement process in clinical practice.

show abstract

Section: ) Error Analysismentioning

confidence: 66%

Section: B Model Performance Benchmarked Against Manual Methods 1) Comparison Of Ctr Mean Errormentioning

confidence: 77%

Section: Deep Learning Approaches To Ctr Estimationmentioning

confidence: 99%

See 1 more Smart Citation

External Validation of Deep Learning Algorithms for Cardiothoracic Ratio Measurement

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Studies have previously shown that machine learning algorithms, namely, SVM, RF (random forest), LR (logistic regression), BPNN (back propagation neural network), and MLP (multilayer perceptron) are utilized successfully for decision-making tools to predict heart disease based on individual information. Various studies revealed the hybrid model merits that achieved good performance in heart disease prediction, namely, RF with a linear model, MLP, Bayes Net (BN), majority voting of NB, and RF and two stacked SVMs, respectively [14]. Kalia Orphanou et al, in the NB classification model, the TARs (Temporal Association Rules) feature is used for diagnosing heart disease.…”

Section: Part 2 Is 'Diastole'mentioning

confidence: 99%

Machine Learning Technology-Based Heart Disease Detection Models

Nagavelli

Samanta

Chakraborty

2022

Journal of Healthcare Engineering

View full text Add to dashboard Cite

At present, a multifaceted clinical disease known as heart failure disease can affect a greater number of people in the world. In the early stages, to evaluate and diagnose the disease of heart failure, cardiac centers and hospitals are heavily based on ECG. The ECG can be considered as a regular tool. Heart disease early detection is a critical concern in healthcare services (HCS). This paper presents the different machine learning technologies based on heart disease detection brief analysis. Firstly, Naïve Bayes with a weighted approach is used for predicting heart disease. The second one, according to the features of frequency domain, time domain, and information theory, is automatic and analyze ischemic heart disease localization/detection. Two classifiers such as support vector machine (SVM) with XGBoost with the best performance are selected for the classification in this method. The third one is the heart failure automatic identification method by using an improved SVM based on the duality optimization scheme also analyzed. Finally, for a clinical decision support system (CDSS), an effective heart disease prediction model (HDPM) is used, which includes density-based spatial clustering of applications with noise (DBSCAN) for outlier detection and elimination, a hybrid synthetic minority over-sampling technique-edited nearest neighbor (SMOTE-ENN) for balancing the training data distribution, and XGBoost for heart disease prediction. Machine learning can be applied in the medical industry for disease diagnosis, detection, and prediction. The major purpose of this paper is to give clinicians a tool to help them diagnose heart problems early on. As a result, it will be easier to treat patients effectively and avoid serious repercussions. This study uses XGBoost to test alternative decision tree classification algorithms in the hopes of improving the accuracy of heart disease diagnosis. In terms of precision, accuracy, f1-measure, and recall as performance parameters above mentioned, four types of machine learning (ML) models are compared.

show abstract

“…This has been accelerated by the availability of large, publicly available clinical imaging datasets [24,12]. Deep learning methods focus on using convolutional neural networks to either assign a binary cardiomegaly diagnosis based on the input image [19], or to use U-Net networks to segment the heart and lungs from the image and estimate the CTR [18,23].…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Classification of Cardiomegaly Using Combined Imaging and Non-imaging ICU Data

Grant

Papież

Parsons

et al. 2021

Medical Image Understanding and Analysis

View full text Add to dashboard Cite

In this paper, we investigate the classification of cardiomegaly using multimodal data, combining imaging data from chest radiography with routinely collected Intensive Care Unit (ICU) data comprising vital sign values, laboratory measurements, and admission metadata. In practice a clinician would assess for the presence of cardiomegaly using a synthesis of multiple sources of data, however, prior machine learning approaches to this task have focused on chest radiographs only. We show that non-imaging ICU data can be used for cardiomegaly classification and propose a novel multimodal network trained simultaneously on both chest radiographs and ICU data. We compare the predictive power of both single-mode approaches with the joint network. We use a subset of data from the publicly available MIMIC-CXR and MIMIC-IV datasets, which contain both chest radiographs and non-imaging ICU data for the same patients. The approach from non-imaging ICU data alone achieves an AUC of 0.684 and the standard chest radiography approach an AUC of 0.840. Our joint model achieves an AUC of 0.880. We conclude that non-imaging ICU data have predictive value for cardiomegaly, and that combining chest radiographs with non-imaging ICU data has the potential to improve model performance for the same subset of patients, with further work required to demonstrate a significant improvement.

show abstract

Cardiomegaly Detection on Chest Radiographs: Segmentation Versus Classification

Cited by 37 publications

References 59 publications

External Validation of Deep Learning Algorithms for Cardiothoracic Ratio Measurement

External Validation of Deep Learning Algorithms for Cardiothoracic Ratio Measurement

Machine Learning Technology-Based Heart Disease Detection Models

Deep Learning Classification of Cardiomegaly Using Combined Imaging and Non-imaging ICU Data

Contact Info

Product

Resources

About