In this chapter, the authors discuss a brief literature review on AI-driven COVID-19 patient screening technologies along with their pros and cons. Next, they discuss the need of human-centric design thinking in AI-driven solutions. Next, in this study, a novel C3IA (CNN-based COVID-19 chest image analysis) tool for automatic COVID-19 detection with multi-class classification (COVID/normal/pneumonia) using raw chest x-ray images is proposed. The authors implemented a two-stream CNN architecture with two pre-trained VGG-16 models, which incorporates both segmented and un-segmented image features. The trained model is tested on more than 500 Indian patient x-ray image datasets and confirmed accuracy of 99% in detecting COVID signatures. Further, in this work, the authors discuss how the design thinking approach is followed in various stages while developing the product to provide a user-friendly efficient real-time COVID-19 chest x-ray image analysis for the common citizen.
TopIntroduction
On December 31, 2019, COVID-19 pandemic eruption in Wuhan, Hubei, China, turned as a rapidly spreader pandemic endangering humankind. Severe acute respiratory syndrome Coronavirus (SARS-CoV) caused severe respiratory disease and death in humans (Khan et. al., 2021). The most common symptoms like dry cough, loss of smell and taste, fever, fatigue, respiratory illness and shortness of breath were experienced by COVID-19 patients. To mitigate the transmission of the novel virus, most of the countries followed lockdowns/curfews with immediate effect along with encouraging various research organizations, higher educational institutes, individuals, scientists, and researchers to work on technological solutions including AI technologies to diminish the COVID-19 pandemic along with discovering effective vaccines (Khan et. al., 2021). Many governments took major steps to accelerate artificial Intelligence (AI) based research in tackling this global health crisis response and healthcare management. In response, AI proved to be crucial technology in building next-generation epidemic preparedness along with a resilient recovery.
Viral nucleic acid detection using the real-time polymerase chain reaction (RT-PCR) and anti-body test are currently the agreed standard method used for COVID-19 diagnosis which are costly, invasive, labor-intensive, time-consuming, diagnostic techniques that require specific instruments. CT Scanning/X-ray imaging modalities has already proven as an adjunctive tool/effective imaging technique for the diagnosis of lung-related diseases like pneumonia and also they does not require any reagents. These technologies can be used for rapid screening of COVID patients as evolution of ground-glass opacification with consolidation in the peripheries and subsequent resolution of the airspaces changes in COVID patients radiological images is seen commonly. Researchers found the symptoms or patterns like right infrahilar airspace opacities, single nodular opacity in the left lower lung region, many irregular opacities in both lungs, ground-glass opacities (GGO), consolidation, and vascular dilation in the lesion, interlobular septal thickening and air bronchogram sign, with or without vascular expansion, Peripheral focal or multifocal GGO affecting both lungs, rounded lung opacities in chest XRay/CT images of various COVID patients (Chakraborty et. al., 2020).
For rapid screening and tracking of unique COVID signatures pertaining to lung irregularities, AI based technological solutions can be employed. With the emergence of machine learning (ML) and deep learning (DL) technologies (subset of AI), the feasibility of AI based data driven solutions is already proven in various domains such as, in product industries, process industries, defence, agriculture, and banking domains. Deep learning (DL) based techniques have proven robust systems for both automated feature engineering and classification of complex imaging data and hence gaining popularity for building automatic diagnosis/assisting tools for clinicians in the medical field.
Recently, many radiology images have been widely used for COVID-19 detection. Chest XRay/CT scan images have been an actively used imaging technique for the diagnosis of lung-related diseases like pneumonia (Tulin et. al., 2020). Tulin et. al., (2020) proposed DarkNet model and reported a classification accuracy of 98.08% for binary classes and 87.02% for multi-class cases. Yicheng et al. (2020) reported the high sensitivity (up to 98%) of chest CT for COVID-19 screening in a series of 51 patients. To speed up the screening, Ophir et al. (2020) employed the DL techniques to detect COVID-19 on CT images. Shi et al. (2021) collected a large-scale COVID-19 CT dataset and developed a ML based method for COVID 19 screening. CT scanning modality suffers from disadvantages like non-portability, high cost, less availability, chance of human to human transmission, and requirement of more technical expertise. Chest Xray is more widely accessible due to its faster imaging time and considerably lower cost than CT, and COVID-19 test kit (Chakraborty et. al., 2020). X-Ray images can be taken at patient place itself/ICU patients; hence they can be convenient mode for later assessments too. So for real-time low cost massive screening of COVID patients, X-Ray modalities can be used in AI based technological solutions.