key: cord-0869892-57sy40ng authors: Panwar, Harsh; Gupta, P.K.; Siddiqui, Mohammad Khubeb; Morales-Menendez, Ruben; Bhardwaj, Prakhar; Singh, Vaishnavi title: A Deep Learning and Grad-CAM based Color Visualization Approach for Fast Detection of COVID-19 Cases using Chest X-ray and CT-Scan Images date: 2020-08-07 journal: Chaos Solitons Fractals DOI: 10.1016/j.chaos.2020.110190 sha: 0b0ac438cfc332c11bae15ca2673a49141592484 doc_id: 869892 cord_uid: 57sy40ng The world is suffering from an existential global health crisis known as the COVID-19 pandemic. Countries like India, Bangladesh, and other developing countries are still having a slow pace in the detection of COVID-19 cases. Therefore, there is an urgent need for fast detection with clear visualization of infection is required using which a suspected patient of COVID-19 could be saved. Therefore, there is an urgent need for fast detection and clear visualization of infection is required. In the recent technological advancements, the fusion of deep learning classifiers and medical images provides more promising results corresponding to traditional RT-PCR testing while making detection and predictions about COVID-19 cases with increased accuracy. In this paper, we have proposed a deep transfer learning algorithm that accelerates the detection of COVID-19 cases by using X-ray and CT-Scan images of the chest. It is because, in COVID-19, initial screening of chest X-ray (CXR) may provide significant information in the detection of suspected COVID-19 cases. We have considered three datasets known as 1) COVID-chest X-ray, 2) SARS-COV-2 CT-scan, and 3) Chest X-Ray Images (Pneumonia). In the obtained results, the proposed deep learning model can detect the COVID-19 positive cases in ≤ 2 seconds which is faster than RT-PCR tests currently being used for detection of COVID-19 cases. We have also established a relationship between COVID-19 patients along with the Pneumonia patients which explores the pattern between Pneumonia and COVID-19 radiology images. In all the experiments, we have used the Grad-CAM based color visualization approach in order to clearly interpretate the detection of radiology images and taking further course of action. 1 Two frailties affect human beings greatly i.e., ill-health and economic 2 slowdown. Unfortunately, this novel coronavirus has brought them to the fore. 3 As we know, the COVID-19 virus targets the lungs of a suspected patient 4 and mutates there promptly. In such a scenario, the infected lungs become 5 inflamed and get filled with fluid. If we perform CT-Scan or X-ray imaging 6 of an infected person then the obtained results show shadowy patches in the 7 lungs called Ground Glass Opacity [1] . Due to the communicable nature, its 8 spread rate is much higher than its prediction or detection rate. 9 During the various experimental findings performed in this paper, it is 10 revealed that the condition of lungs infected with COVID-19 often relates to 11 another common lung infection known as Pneumonia. Usually, pneumonia is 12 an infection caused by either a virus, bacteria, or fungi. Similar to 13 pneumonia can also be life-threatening for the age group below 2 and people 14 above 65. It is contagious just like COVID-19 with several similar symptoms. 15 However, COVID-19 has proved itself to be more fatal than Pneumonia, as it 16 can lead on to cause Acute Respiratory Distress Syndrome (ARDS). In other 17 words, we can say that it happens as a result of progressing Pneumonia in 18 the lungs. 19 In the absence of an intelligent diagnosis method, there is a great re-20 quirement for fast and accurate detection of COVID-19 suspected patients. 21 However, the contemporary techniques available for the detection of this raging The main contributions of the paper are listed here: • We have proposed a new deep transfer learning algorithm and tested it 77 on three different radiology datasets for faster detection of COVID-19. • With the usage of deep learning and its advantage, it is assumed that 79 proposed model is faster than the traditionally used RT-PCR testing 80 kit. • We also discover through the experimental by our proposed algorithm 82 that there is strong relevance between Pulmonary diseases like Pneumo-83 nia and COVID-19. • Grad-CAM analysis has been performed over obtained results provide 85 the coloured visualization of the regions of lungs infected by the COVID- 86 19 in both CT-scan and X-ray images. • We also make the usage of false positive cases e.g., Pneumonia and 88 consider them in COVID-19 suspected category. • Efficacy of proposed model is more accurate on CT-scan compare to 90 CXR images. where z [1] is the current layer, a [0] is the first or input layer, W [1] represents 187 the weights for the first layer and b [1] is the bias. For VGG19 Conv Layer [24], c . The final layer is shown in Eq. 3. Or, Where i, j and k corresponds to row, column and channel for z [1] respec- it was able to achieve 75.2% top-1 and 92.5% top-5 accuracy. Therefore, we 247 have considered it as one of the base models for the proposed study. Here, the last 5 layers include an Average Pooling 2D layer which also 262 results in dimension reduction by reckoning the average values of each region. The average pooling layer is followed by a flatten layer that creates a single For evaluating the model we have calculated the precision, recall, F-284 measure score and accuracy using the confusion matrix as shown in Table 5 . where n represents the n th class. As, we have considered the binary image 315 classification therefore the value of class will be n = 2. For this, Here, we can 316 select any required datasets i.e δ 1 or δ 2 to the input variables in Algorithm 1. However, for multiclass image classification the value of n should be ≥ 2 . The main steps of the proposed algorithm are as follows: hyper-parameters as shown in Table 1 using which the best accuracy 337 for the proposed model has been obtained. where, y denotes the true value, and p denotes the probability predicted 352 by model. Fig. 8(a) . The primary dataset is divided into three sub-368 datasets i.e. training, validation, and testing sets. In train and test split, 369 we have used 80% of images for training, and the remaining 20% for testing Fig. 10(a) and 10(b) . The recall, precision, and F1-score are also 393 mentioned in the Table 5 . From Table 5 , we can find that proposed model has The count of images of Pneumonia and COVID-19 are highlighted in 399 Fig. 8(b) . The objective of this experiment is explore the relationship between Here, values obtained after Experiment-2 for training accuracy, validating These results are shown in Fig. 10 (c) and 10(d). Here, we can find the clear Through the confusion matrix presented in Table 4 Early ct features and temporal lung changes in covid-19 pneumonia in 473 wuhan, china Deep learning in medical image analysis Coronavirus pandemic: A predictive analysis of 481 the peak outbreak epidemic in South Africa, Turkey, and Brazil Role of intelligent computing in covid-19 prognosis: 484 A state-of-the-art review Hus-486 sain, K. Khatoon, S. Ahmad, Correlation between temperature and 487 covid-19 (suspected, confirmed and death) cases based on machine learn-488 ing analysis Deep learning system to screen coronavirus disease 2019 491 pneumonia Rapid ai development cycle for the 494 coronavirus (covid-19) pandemic: Initial results for automated detection 495 & patient monitoring using deep learning ct image analysis Analysis and forecast of covid-19 spreading in 498 china, italy and france Finding covid-19 500 from chest x-rays using deep learning on a small dataset Artificial intelligence and machine learning to fight covid-19 Deep learning system to screen 505 coronavirus disease 2019 pneumonia Application of Deep Learning for Fast detection of COVID-19 in X-rays 508 using nCOVnet Deep transfer learning-based covid-19 510 prediction using chest x-rays, medRxiv A transfer learning method for pneumonia classification and 513 visualization Covid-19: automatic detection 515 from x-ray images utilizing transfer learning with convolutional neural 516 networks Covid-19 detection using deep learning 518 models to exploit social mimic optimization and structured chest x-ray 519 images using fuzzy color and stacking approaches Automated detection of COVID-19 cases using deep neural 523 networks with x-ray images Covid-19 image data collection CT-scan dataset: A large dataset of real patients CT scans for SARS-529 CoV-2 identification, medRxiv Labeled Optical Coherence Tomog-531 raphy (OCT) and Chest X-Ray Images for Classification Chest X-Ray Images (Pneumonia) Deep learning Very Deep Convolutional Networks for 540 Large-Scale Image Recognition Convolutional Net-543 works An introduction to convolutional neural networks Visual explanations from deep networks via gradient-based 550 localization Analyzing performance of 553 classification techniques in detecting epileptic seizure Conference on Advanced Data Mining and Applications