Four Types of Multiclass Frameworks for Pneumonia Classification and Its Validation in X-ray Scans Using Seven Types of Deep Learning Artificial Intelligence Models

Abstract

Background and Motivation: The novel coronavirus causing COVID-19 is exceptionally contagious, highly mutative, decimating human health and life, as well as the global economy, by consistent evolution of new pernicious variants and outbreaks. The reverse transcriptase polymerase chain reaction currently used for diagnosis has major limitations. Furthermore, the multiclass lung classification X-ray systems having viral, bacterial, and tubercular classes—including COVID-19—are not reliable. Thus, there is a need for a robust, fast, cost-effective, and easily available diagnostic method. Method: Artificial intelligence (AI) has been shown to revolutionize all walks of life, particularly medical imaging. This study proposes a deep learning AI-based automatic multiclass detection and classification of pneumonia from chest X-ray images that are readily available and highly cost-effective. The study has designed and applied seven highly efficient pre-trained convolutional neural networks—namely, VGG16, VGG19, DenseNet201, Xception, InceptionV3, NasnetMobile, and ResNet152—for classification of up to five classes of pneumonia. Results: The database consisted of 18,603 scans with two, three, and five classes. The best results were using DenseNet201, VGG16, and VGG16, respectively having accuracies of 99.84%, 96.7%, 92.67%; sensitivity of 99.84%, 96.63%, 92.70%; specificity of 99.84, 96.63%, 92.41%; and AUC of 1.0, 0.97, 0.92 (p < 0.0001 for all), respectively. Our system outperformed existing methods by 1.2% for the five-class model. The online system takes <1 s while demonstrating reliability and stability. Conclusions: Deep learning AI is a powerful paradigm for multiclass pneumonia classification.

Keywords: COVID-19; Omicron; chest X-rays; convolutional neural network; deep learning; transfer learning.

Figure 1

Overall schematic diagram of the proposed method for multiclass scenario using seven deep learning models.

Figure 2

Sample chest X-ray images from each class.

Figure 3

(a) VGG16 transfer learning architecture. (b) VGG19 transfer learning architecture. (c) DenseNet201 transfer learning architecture. (d) Xception transfer learning architecture. (e) InceptionV3 transfer learning architecture. (f) NasnetMobile transfer learning architecture. (g) ResNet152 transfer leaning architecture.

Figure 3

(a) VGG16 transfer learning architecture. (b) VGG19 transfer learning architecture. (c) DenseNet201 transfer learning architecture. (d) Xception transfer learning architecture. (e) InceptionV3 transfer learning architecture. (f) NasnetMobile transfer learning architecture. (g) ResNet152 transfer leaning architecture.

Figure 3

(a) VGG16 transfer learning architecture. (b) VGG19 transfer learning architecture. (c) DenseNet201 transfer learning architecture. (d) Xception transfer learning architecture. (e) InceptionV3 transfer learning architecture. (f) NasnetMobile transfer learning architecture. (g) ResNet152 transfer leaning architecture.

Figure 4

Training and validation accuracy of best performing VGG16 network for COVID-19 and normal class.

Figure 5

Training and validation loss of best performing VGG16 network for COVID-19 and normal class.

Figure 6

Confusion matrix for the classification into COVID-19 and normal by VGG16.

Figure 7

Training and validation accuracy of best performing NasnetMobile model for COVID-19 and viral pneumonia class.

Figure 8

Training and validation loss of best performing NasnetMobile model for COVID-19 and viral pneumonia class.

Figure 9

Confusion matrix for the classification into COVID-19 and viral pneumonia by NasnetMobile.

Figure 10

Training and validation accuracy of best performing DenseNet201 model for COVID-19 and bacterial pneumonia class.

Figure 11

Training and validation loss of best performing DenseNet201 model for COVID-19 and bacterial pneumonia class.

Figure 12

Confusion matrix for the classification into COVID-19 and bacterial pneumonia by DenseNet201.

Figure 13

Training and validation accuracy of best performing VGG16 model for COVID-19 and tuberculosis class.

Figure 14

Training and validation loss of best performing VGG16 model for COVID-19 and tuberculosis class.

Figure 15

Confusion matrix for the classification into COVID-19 and tuberculosis by VGG16.

Figure 16

Training and validation accuracy of best performing VGG16 model for three-class experiment.

Figure 17

Training and validation loss of best performing VGG16 model for three-class experiment.

Figure 18

Confusion matrix for three-class classification by VGG16.

Figure 19

Training and validation accuracy of best performing VGG16 model for five-class.

Figure 20

Training and validation loss of best performing VGG16 model for five-class.

Figure 21

Confusion matrix for five-class classification by VGG16.

Figure 22

ROC curves and AUC values for two-class experiments: (a) COVID-19 and normal by VGG16; (b) COVID-19 and viral pneumonia by NasnetMobile; (c) COVID-19 and bacterial pneumonia by Densenet201; (d) COVID-19 and tuberculosis by VGG16.

Figure 23

ROC curves and AUC values for three-class experiment by VGG16.

Figure 24

ROC curves and AUC values for five-class experiment by VGG16.

Figure 25

Sample images from the first eight classes of Faces95 database.

Figure 26

Training and validation accuracy of best performing VGG16 model for Faces95 images.

Figure 27

Training and validation loss of best performing VGG16 model for Faces95 images.