Article Preview
Top1. Introduction
Owing to environmental changes, global warming, behaviours and other factors, the health effect of the disease is growing quickly. There has been a growing risk of health issues. Due to a COPD, commonly caused by pollution and cigarettes, about 3.4 million died in 2016, while 400,000 died of asthma (Underwood, 2017). The risk of lung diseases is enormous, particularly in developing and low-income countries, where millions of people are facing poverty and air pollution (Chowdhury, Mayilvahananan, & Govindaraj, 2019). The WHO reports that the households suffer from air pollution, including asthma and pneumonia, contribute to more than four million premature deaths each year. Consequently it is important to take access to air pollution and carbon emissions. Efficient diagnostic systems to help detect pulmonary diseases are also essential. A novel 2019 coronavirus disease has caused severe pulmonary damage and air problems since late December 2019 (COVID-19). In addition to this, pneumonia may be caused by a form of lung disease or by other viral or bacterial infections due to the COVID-19 causative virus (Yang et al., 2020).
Early detection of lung diseases is therefore increasingly necessary. Machine learning and deep learning will play a crucial role in this (Nair & Bhagat, 2019a). Recently digital technology has gained global significance. This study will help physicians and other researchers detect lung diseases using a deep learning. As datasets are several X-ray images. The device presented can also help diagnose diseases more accurately, saving many vulnerable people and reducing the number of illnesses. The health system has not been developed in part due to population growth (Dickman et al., 2017). Many researchers carried out research related to predicting X-ray picture diagnostic information by machine learning systems (Chowdhury et al., 2020) (Chandra et al., 2016). This is high time to address this complication, since there is no constraint on the control of machines and a huge number of data. This approach could minimise medical costs by extending computer science for healthcare and medical research projects. For implementation NIH X-ray image datasets as shown in figure 1 are downloaded from Kaggle (Gu et al., 2018; Kong et al., 2020), a truly open source platform.
Figure 1. X-ray image (Source: Kaggle)
With the technology of deep learning, a huge amount of data can be analysed and understood more easily. One area where CNN-based models are widely used is medicine. However, a variety of limitations exist in CNN architectures. Another constraint on the prediction system is the pooling number of neurons. Max pooling is designed to migrate the most necessary data to the next higher layer of the neural network. This will cause information in the data to be lost, making the data unreadable in other layers. Current CNN models cannot effectively show the relationship between parts and wholes. To solve the drawbacks of CNNs, Capsule Networks method has been introduced in 2017.
In this paper, the new hybrid algorithm is introduced and effective in classifying pulmonary disease in the above-mentioned data collection. The main aim of this research is to build this new hybrid deep learning algorithm for the prediction of pulmonary X-ray disease. Section 2 describes specifically related papers on the classification and classification of lung X-ray images or lung module. Section 3 discusses the proposed methodology and section 4 presents a result analysis of the data set. Section 5 concludes the paper.