Healthcare Informatics for Fighting COVID-19 and Future Epidemics

Healthcare Informatics for Fighting COVID-19 and Future Epidemics

Chapitre 16 : Explainable Deep Learning for Covid-19 Detection Using Chest X-ray and CT-Scan Images
Authors : Sidi Ahmed Mahmoudi, Sédrick Stassin, Mostafa El Habib Daho, Xavier Lessage, Saïd Mahmoudi

Chapitre 21 : Chest X-ray Images Analysis with Deep Convolutional Neural Networks (CNN) for COVID-19 Detection
Authors : Xavier Lessage Saïd Mahmoudi, Sidi Ahmed Mahmoudi, Sohaib Laraba, Olivier Debauche, Mohammed Amin Belarbi

Date: 12 mars 2021

Publication: Publications scientifiques 


Science des données 

Domaine: Santé 

Auteur : Xavier Lessage

Chapitre 16 - Explainable Deep Learning for Covid-19 Detection Using Chest X-ray and CT-Scan Images

Recently, Artificial Intelligence (AI) and more particularly Deep Learning (DL) applications gained significant importance in several domains such as computer vision, robotics, medical imaging, etc. Despite the excellent results of AI models, in terms of precision and performance, their decisions are not always interpretable and explainable, which makes from them a black box. Since May 2018, the general data protection regulation (GDPR) requires a right of explanation for the output of an algorithm, which is necessary and justified for several examples such as autonomous cars and computer-aided diagnosis (CAD) systems. As a result, a high interest in terms of research has been given recently to the domain of Explainable Artificial Intelligence (XAI). In this book chapter, we propose an approach for explaining Deep Learning algorithms when applied to image classification and segmentation. The proposed approach allows to provide the most appropriate explanation method and the most accurate and explainable DL model. As a use case, we applied our approach for explaining DL models used Covid-19 image classification and segmentation with two modalities : X-ray and CT-scan images. Experimental results showed the interest of our explanation approach within three facts : (1) identification of the most interpretable DL model ; (2) measurement of positive and negative contributions of input parameters (image pixels) in the decision of DL models ; (3) detection of data (training and validation datasets) biases, where the deep neural networks are focusing on image regions that are not supposed to be important. The provided explanations were evaluated by doctors and physicians who confirmed the accuracy of our results.

Explainable artificial intelligence - XAI - Explainable deep learning - Deep learning - Medical imaging - Covid-19 detection - Chest X-ray and CT images - Grad-CAM - LRP Occlusion - Bias detection - Images classification

Chapitre 21- Chest X-ray Images Analysis with Deep Convolutional Neural Networks (CNN) for COVID-19 Detection

The latest advances of deep learning and particularly convolutional neural networks (CNNs) have proven more than once their high accuracy in disease detection. In this chapter, we propose a new deep learning-based approach for COVID-19 detection from chest X-ray images. The proposed approach applies, in an efficient way, the techniques of transfer learning and fine-tuning from pre-trained CNN models (InceptionV3, VGG16, MobileNet, EfficientNet, etc.). The dataset used for our experiments has three classes : normal, COVID-19, and other pathologies. The dataset is split into three sub-sets as follows : 70% for training, 15% for validation, and 15% for the final test. To avoid underfitting or overfitting problems during the training process, we apply regularization techniques (L1 & L2 regularizations, dropout, data augmentation, early stopping, cross-validation, etc.), which help in learning and providing a generalizable solution. As a result, we demonstrate the high efficiency of the proposed CNNs for the detection of COVID-19 from chest X-ray images. A comparison of different architectures shows that VGG16 and MobileNet provide the highest scores : 97.5% and 99.3% of accuracy respectively, 98.7% and 99.3% of sensitivity respectively. In addition, both models provide the scores of 96.3% and 99.2% respectively for specificity. The proposed solution is deployed in the cloud to provide high availability in real time, thanks to a responsive website, and this without the need to download, install, and configure the required libraries.

Chest X-ray analysis - Classification - Convolutional neural networks