COVID-19 was a global disaster for humans in 2020. It actually causes
breathing problems, sneezing, and coughing. Early detection is essential to minimize
the viral effects. RT-PCR test is not suitable against the rapid spread of coronavirus
since it takes time to detect positive cases. Moreover, COVID-19 is a transmittable
disease; therefore, there is always a chance of spreading the disease by RT-PCR
method. Nowadays, radiographic images have already proven to be popular and
alternate diagnostic tools in the context of disease prediction. Several research papers
have already proposed different techniques using deep learning or CNN methods or
transfer learning methods with radiographic images for automatic screening of
COVID-19 disease. Unfortunately, most of these existing screening methods have
either worked on small datasets or have produced the results with low accuracy. In this
paper, an IoT-enabled automated screening model has been presented to classify the
COVID-19 cases with the CNN-LSTM+Capsule network. The LSTM-Capsule network
model not only diagnose the COVID-19 cases quickly but also maintains low hardware
costs. The proposed model has worked on 280 COVID-positive and 290 normal
patients’ images taken from two benchmark datasets and tried to find out the spreading
tendency of the virus by getting some realistic data from IoT sensor devices in the
affected zone. The proposed model achieves an accuracy of 97%. Moreover, crossvalidation is applied in this model to avoid over-fitting. This paper concludes that the
proposed model gives an insight into the CNN-LSTM+Capsule network working for
COVID-19 detection that helps get richer feature mapping from the radiographic
images and further classifies the COVID-19 cases from the normal ones effectively.
Moreover, the LSTM-Capsule network model has outperformed relative to other
related works in the early identification of COVID-19 cases.
Keywords: COVID-19, Capsule network, Computer tomography, CNN-LSTM, IoT, X-ray.
