Exploration of Interpretability Techniques for Deep COVID-19 Classification using Chest X-ray Images

Chatterjee, Soumick; Saad, Fatima; Sarasaen, Chompunuch; Ghosh, Suhita; Khatun, Rupali; Radeva, Petia; Rose, Georg; Stober, Sebastian; Speck, Oliver; Nürnberger, Andreas

doi:10.48550/arxiv.2006.02570

Cited by 5 publications

(4 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This paper presents the TorchEsegeta framework, which integrates various interpretability and explainability techniques available in different libraries and extends these techniques for segmentation models. It is noteworthy that the development of this pipeline started with the exploration of various interpretability techniques for classifying COVID-19 and other types of pneumonia [10]. An initial pipeline was developed under that project for classification models but only for 2D images.…”

Section: Methodsmentioning

confidence: 99%

TorchEsegeta: Framework for Interpretability and Explainability of Image-Based Deep Learning Models

et al. 2022

View full text Add to dashboard Cite

Clinicians are often very sceptical about applying automatic image processing approaches, especially deep learning-based methods, in practice. One main reason for this is the black-box nature of these approaches and the inherent problem of missing insights of the automatically derived decisions. In order to increase trust in these methods, this paper presents approaches that help to interpret and explain the results of deep learning algorithms by depicting the anatomical areas that influence the decision of the algorithm most. Moreover, this research presents a unified framework, TorchEsegeta, for applying various interpretability and explainability techniques for deep learning models and generates visual interpretations and explanations for clinicians to corroborate their clinical findings. In addition, this will aid in gaining confidence in such methods. The framework builds on existing interpretability and explainability techniques that are currently focusing on classification models, extending them to segmentation tasks. In addition, these methods have been adapted to 3D models for volumetric analysis. The proposed framework provides methods to quantitatively compare visual explanations using infidelity and sensitivity metrics. This framework can be used by data scientists to perform post hoc interpretations and explanations of their models, develop more explainable tools, and present the findings to clinicians to increase their faith in such models. The proposed framework was evaluated based on a use case scenario of vessel segmentation models trained on Time-of-Flight (TOF) Magnetic Resonance Angiogram (MRA) images of the human brain. Quantitative and qualitative results of a comparative study of different models and interpretability methods are presented. Furthermore, this paper provides an extensive overview of several existing interpretability and explainability methods.

show abstract

Section: Methodsmentioning

confidence: 99%

TorchEsegeta: Framework for Interpretability and Explainability of Image-Based Deep Learning Models

et al. 2022

View full text Add to dashboard Cite

show abstract

“…This model's network layers have the most information flowing through them, which makes it easier to extract the best characteristics. Yes Yes [29] Yes Yes [30] Yes [31] Yes [32] Yes [33] Yes Yes Yes [34] Yes Yes [35] Yes Yes Yes [36] Yes Yes [24] Yes Yes Yes [25] Yes Yes Yes [37] Yes [38] Yes Yes Yes [39] Yes Yes [39] Yes Yes [40] Yes Yes [41] Yes Yes [42] Yes Yes…”

Section: Data Segmentationmentioning

confidence: 99%

COVID-19 Detection by Medical Images with Pretrained Transfer Learning-based Model using CNN: A Systematic Review

Patel¹,

Khan²

2022

2022 IEEE International Conference on Current Development in Engineering and Technology (CCET)

View full text Add to dashboard Cite

A new coronavirus that caused the Covid-19 sickness, has already elevated the threat to humans. The virus is quickly spreading around the planet. Therefore, in order to detect sick individuals and stop the infection from spreading, it is vital that we develop fast diagnostic tests. The advancement of machine learning would make it possible to implement pre-ventative actions as soon as possible by enabling early detection of Covid19. However, insufficient sample sizes, particularly chest X-ray pictures, has made it more challenging to diagnose this ailment. In this study, we examined a number of these recently created transfer learning-based CNN models that can identify COVID-19 in lung CT or images of X-ray to diagnose Covid-19 using images of X-ray. We gathered data on the research resources that are readily available. We looked into and examined datasets, pre-processing methods, segmentation approaches, ex-traction of features, classification, and experimentation outcomes that could be useful for determining future research paths in the area of applying transfer learning based CNN models to diagnose COVID-19 disease. We have analyzed various models such as ResNet50, DenseNet-21, VGG-16, ImageNet, and some hybrid models and evaluated their performance matrix with a particular set of data used in their research work. Additionally, in order for a model to perform at its best, it is observed that there aren't enough data sets of COVID-19-infected individuals. This calls for augmentation, segmentation, and domain adaptation in transfer learning.

show abstract

“…One of the main contributions of this research work is the TorchEsegeta framework which integrates various interpretability and explainability techniques available in different libraries and extends these techniques for segmentation models. It is noteworthy that the development of this pipeline started with the exploration of various interpretability techniques for classifying COVID-19 and other types of pneumoniae (Chatterjee et al, 2020b). An initial pipeline was developed under that project for classification models, but only for 2D images.…”

Section: Architecture Of Torchesegetamentioning

confidence: 99%

TorchEsegeta: Framework for Interpretability and Explainability of Image-based Deep Learning Models

Chatterjee¹,

Das²,

Mandal³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Clinicians are often very sceptical about applying automatic image processing approaches, especially deep learning based methods, in practice. One main reason for this is the black-box nature of these approaches and the inherent problem of missing insights of the automatically derived decisions. In order to increase trust in these methods, this paper presents approaches that help to interpret and explain the results of deep learning algorithms by depicting the anatomical areas which influence the decision of the algorithm most. Moreover, this research presents a unified framework, TorchEsegeta, for applying various interpretability and explainability techniques for deep learning models and generate visual interpretations and explanations for clinicians to corroborate their clinical findings. In addition, this will aid in gaining confidence in such methods. The framework builds on existing interpretability and explainability techniques that are currently focusing on classification models, extending them to segmentation tasks. In addition, these methods have been adapted to 3D models for volumetric analysis. The proposed framework provides methods to quantitatively compare visual explanations using infidelity and sensitivity metrics. This framework can be used by data scientists to perform posthoc interpretations and explanations of their models, develop more explainable tools and present the findings to clinicians to increase their faith in such models. The proposed framework was evaluated based on a use case scenario of vessel segmentation models trained on Time-of-fight (TOF) Magnetic Resonance Angiogram (MRA) images of the human brain. Quantitative and qualitative results of a comparative study of different models and interpretability methods are presented. Furthermore, this paper provides an extensive overview of several existing interpretability and explainability methods.

show abstract

Exploration of Interpretability Techniques for Deep COVID-19 Classification using Chest X-ray Images

Cited by 5 publications

References 33 publications

TorchEsegeta: Framework for Interpretability and Explainability of Image-Based Deep Learning Models

TorchEsegeta: Framework for Interpretability and Explainability of Image-Based Deep Learning Models

COVID-19 Detection by Medical Images with Pretrained Transfer Learning-based Model using CNN: A Systematic Review

TorchEsegeta: Framework for Interpretability and Explainability of Image-based Deep Learning Models

Contact Info

Product

Resources

About