[This article belongs to Volume - 55, Issue - 01, 2023]
Gongcheng Kexue Yu Jishu/Advanced Engineering Science
Journal ID : AES-30-03-2023-046

1A. Phani Sridhar, 2 Dr.P.V. Lakshmi

Abstract :

Background: Novel Corona Virus is increasing day by day; it is needed to identify the formal techniques and innovative trendy approaches. The responsive effort in short times is needed. Objective: The aim of this study was to consequently, display things that recognize COVID-19 pneumonia accurately. This examine work is pointing to particularly offer assistance with the conclusion of COVID-19. Methods: The proposed work is centred on the headway of an AI-based examination of CT pictures of the cutting edge crown infection. The proposed system classifies CT pictures utilizing balanced Resnet appear name ResNext. The Convolutional layers of particular sizes were utilized in each of the confined ways of Resnet illustrated. In Resnext, 32 channels are bound together at the same bottleneck and convolve them. This made it conceivable to perform indistinguishable changes to gather convolution in 32 bunches, which compares to the initial 32 courses. The proposed show effectively separated between viral pneumonia and COVID-19 influenced lung CT pictures. The accuracies of DenseNet, mobileNet and VggNet are 90.91,75.24 and 35.75 respectively for testing as shown in table 4. Results: False positives are identified among normal images, Covid effected images and viral pneumonia images. Also the validation loss and validation accuracy of the training process is to be processed and observed. The training loss is also calculated and observed. The training accuracy and validation accuracy both reached 100%. The final testing accuracy of the proposed model is 100%. The performance comparison of existing methods with the proposed method is also observed. The proposed method obtained better classification accuracy when compared to the existing deep learning models DenseNet, mobileNet and VggNet. The accuracies of DenseNet, mobileNet and VggNet are 89.3, 72.72 and 30.30 respectively for training. Conclusion: The clinical execution of the PCR test for COVID-19 pneumonia is not accurate. This paper points to assist make stride the AI examination stage for COVID-19 pneumonia investigate and make strides the precision of AI choice and judgment. The proposed demonstrate gotten an exactness of 100% in dispensing with wrong positives delivering accurate COVID-19 detection.