Multicenter Validation of Convolutional Neural Networks for Automated Detection of Cardiomegaly on Chest Radiographs

Cárdenas, Diego J.; Ferreira, J.A.M.; Moreno, Ramón; Rebelo, M.S.; Krieger, José Eduardo; Gutierrez, Marco Antônio

doi:10.5753/sbcas.2020.11512

Cited by 6 publications

(7 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Subsequently, we applied a data augmentation technique to compensate for the reduced number of samples. Our preprocessing steps are based on previous works [7, 11] for cardiomegaly classification. In order to improve the efficiency of the training step and reduce the computational cost, we used a previously developed and validated model [20] that segments the lungs, to be able to create a binary mask of the chest cavity region, using a fine-tuned UNet-based CNN.…”

Section: Methodsmentioning

confidence: 99%

Explainable AI in Deep Learning-based Detection of Aortic Elongation on Chest X-ray Images

Ribeiro,

Cardenas,

Dias

et al. 2023

Preprint

View full text Add to dashboard Cite

Aortic Elongation can result from age-related changes, congenital factors, aneurysms, or conditions affecting blood vessel elasticity. It is associated with cardiovascular diseases and severe complications like aortic aneurysms and dissection. In this work, we evaluated the performance of deep learning models (DenseNet and EfficientNet) for aortic elongation detection based on transfer learning and fine-tunning techniques using Chest X-Rays (CXR) as input. DenseNet achieved higher accuracy (84.7% ± 0.7), precision (75.6% ± 1.3), sensitivity (88.7% ± 2.7), specificity (82.3% ± 1.6), F1-score (81.6% ± 1.0), and AUROC (93.1% ± 0.4) than EfficientNet. To gain insights into the decision-making process of the deep learning models, we employed Grad-CAM and LIME explainability methods. Through these techniques, we were able to successfully identify the expected location of aortic elongation in the x-ray images. Moreover, we used the pixel-flipping method to assess quantitatively the interpretations providing valuable insights into models behavior. By incorporating explainable AI techniques, we enhanced the interpretability and understanding of the models' decisions. This approach holds promise for aiding clinicians in timely and accurate diagnosis, potentially improving patient outcomes in clinical practice.

show abstract

Section: Methodsmentioning

confidence: 99%

Explainable AI in Deep Learning-based Detection of Aortic Elongation on Chest X-ray Images

Ribeiro,

Cardenas,

Dias

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The features consisting of the output specific to the top fully connected layer of each single model are extracted and concatenated, before being fed into subsequent fully connected layers to generate the final output. Similarly, different authors (Bougias et al, 2020 ; Cardenas et al, 2020 ) evaluate several pre-trained DNNs for the detection of cardiomegaly based on transfer learning approaches, including such models as VGG16 and VGG19 (Simonyan and Zisserman, 2015 ), MobileNet (Howard et al, 2017 ), DenseNet121 (Huang et al, 2017 ), and EfficientNetB2 (Tan and Le, 2019 ). Bougias et al ( 2020 ) replace the top fully connected layers of each pre-trained model by a logistic regression classifier before proceeding with the optimization of the resulting architecture.…”

Section: Related Workmentioning

confidence: 99%

“…Bougias et al ( 2020 ) replace the top fully connected layers of each pre-trained model by a logistic regression classifier before proceeding with the optimization of the resulting architecture. Meanwhile, Cardenas et al ( 2020 ) replace the entire fully connected layers of each pre-trained model by a customized 3-layer multilayer perceptron (MLP). Uniquely the weights specific to the MLP are subsequently optimized during the training process.…”

Section: Related Workmentioning

confidence: 99%

Unsupervised domain adaptation for the detection of cardiomegaly in cross-domain chest X-ray images

Thiam¹,

Lausser²,

Kloth³

et al. 2023

Front. Artif. Intell.

View full text Add to dashboard Cite

In recent years, several deep learning approaches have been successfully applied in the field of medical image analysis. More specifically, different deep neural network architectures have been proposed and assessed for the detection of various pathologies based on chest X-ray images. While the performed assessments have shown very promising results, most of them consist in training and evaluating the performance of the proposed approaches on a single data set. However, the generalization of such models is quite limited in a cross-domain setting, since a significant performance degradation can be observed when these models are evaluated on data sets stemming from different medical centers or recorded under different protocols. The performance degradation is mostly caused by the domain shift between the training set and the evaluation set. To alleviate this problem, different unsupervised domain adaptation approaches are proposed and evaluated in the current work, for the detection of cardiomegaly based on chest X-ray images, in a cross-domain setting. The proposed approaches generate domain invariant feature representations by adapting the parameters of a model optimized on a large set of labeled samples, to a set of unlabeled images stemming from a different data set. The performed evaluation points to the effectiveness of the proposed approaches, since the adapted models outperform optimized models which are directly applied to the evaluation sets without any form of domain adaptation.

show abstract

“…Cardiomegaly, a pathology frequently detected in chest x-rays (CXR), is a medical disorder in which a patient's heart enlarges temporarily or permanently depending on the situation. This enlargement is typically a clinical manifestation of another pathogenic condition, such as chamber dilation, ventricular hypertrophy, or pericardial effusion [Daines et al 2021], possibly resulting in heart failure, cardiac arrest, and sudden death [Cardenas et al 2020]. Routine chest radiographs showing cardiomegaly have significant clinical implications, since they help doctors decide whether additional investigation is necessary [Daines et al 2021].…”

Section: Introductionmentioning

confidence: 99%

Interpretable Deep Learning Model For Cardiomegaly Detection with Chest X-ray Images

Ribeiro,

Cardenas,

Krieger

et al. 2023

Anais Do XXIII Simpósio Brasileiro De Computação Aplicada À Saúde (SBCAS 2023)

View full text Add to dashboard Cite

Cardiomegaly is a medical disorder characterized by an enlargement of the heart. Many works propose to automatically detect cardiomegaly through chest X-rays. However, most of them are based on deep learning models, known for their lack of interpretability. This work propose a deep learning model for the detection of cardiomegaly based on chest x-rays images and the qualitative assessment of three known local explainable methods, i.e., Grad-CAM, LIME and SHAP. Our model achieved Acc, Prec, Se, Spe, F1-score and AUROC of 91.8±0.7%, 74.0±2.7%, 87.0±5.5%, 92.9±1.2%, 79.8±1.9%, and 90.0±0.7%, respectively. Moreover, except for the SHAP method, our interpretable methods were able to pinpoint the expected location for cardiomegaly. However, Grad-CAM method showed faster computational time than LIME and SHAP.

show abstract

Multicenter Validation of Convolutional Neural Networks for Automated Detection of Cardiomegaly on Chest Radiographs

Cited by 6 publications

References 16 publications

Explainable AI in Deep Learning-based Detection of Aortic Elongation on Chest X-ray Images

Explainable AI in Deep Learning-based Detection of Aortic Elongation on Chest X-ray Images

Unsupervised domain adaptation for the detection of cardiomegaly in cross-domain chest X-ray images

Interpretable Deep Learning Model For Cardiomegaly Detection with Chest X-ray Images

Contact Info

Product

Resources

About