key: cord-0935664-4rplgixe authors: Adwibowo, A. title: Machine learning and AI aided tool to differentiate COVID 19 and non-COVID 19 lung CXR date: 2020-08-21 journal: nan DOI: 10.1101/2020.08.18.20175521 sha: 0c01f0b159220dabd60f737d304ef1a682a84263 doc_id: 935664 cord_uid: 4rplgixe One of the main challenges in dealing with the current COVID 19 pandemic is how to detect and distinguish between the COVID 19 and non COVID 19 cases. This problem arises since COVID 19 symptoms resemble with other cases. One of the golden standards is by examining the lung using the chest X ray radiograph (CXR). Currently there is growing COVID 19 cases followed by the CXR images waiting to be analyzed and this may outnumber the health capacity. Learning from that current situation and to fulfill the demand for CXRs analysis, a novel solution is required. The tool is expected can detect and distinguish the COVID 19 case lung rely on CXR. Respectively, this study aims to propose the use of AI and machine learning aided tool to distinguish the COVID 19 and non COVID 19 cases based on the CXR lung image. The compared non COVID 19 CXR cases in this study include normal (healthy), influenza A, tuberculosis, and active smoker. The results confirm that the machine learning tool is able to distinguish the COVID 19 CXR lungs based on lung consolidation. Moreover, the tool is also able to recognize an abnormality of COVID 19 lung in the form of patchy ground glass opacity. Important approach to deal with the COVID 19 is to find the positive confirmed cases. However, recognizing COVID 19 cases within a community is still becoming a challenge. Related to that, a wide array of COVID 19 medical diagnosis has been developed to support the detection of the COVID 19 cases cases based on the CXR image. The methodology used in this study is . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 21, 2020. . https://doi.org/10.1101/2020.08.18.20175521 doi: medRxiv preprint The developed machine learning aided tool is Besides measure the difference between COVID 19 and non COVID 19 lungs, the machine learning is able to detect and mark the abnormalities. Figure 1 shows the abnormality of COVID 19 right lung in the form of patchy ground glass opacity that has been marked. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 21, 2020. . This study In this study, a CXR of COVID 19 and non COVID 19 cases were used to be analyzed using the AI and machine learning tools. Mangal et al (2020) confirm that the use of CXR has several advantages. First, CXR is more accessible, widespread, and low cost compared to CT scan. In the operational context, CXR is available as portable machine and this enables to conduct X ray test within an isolation ward and reduce the requirement of Personal Protective Equipment. The CXR is also has reliable accuracy to detect COVID 19. Ai et al (2020) reported that sensitivity of CXR was 97% based on positive RT PCR results. A rapid COVID 19 cases worldwide has demanded a robust and versatile detection tools. Based on the golden standard, the developed tools recently were relied on CXR (Cicero et al 2017) and supported with the CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 21, 2020 . . https://doi.org/10.1101 analysis performed by AI and machine learning (Chowdhury et al 2020) . A lot of studies have been done in this field and noticeable trend of study using AI and machine learning aided medical diagnosis has been observed in the last few years. The trend is growing following recent COVID 19 cases. The result derived from this study is also comparable with the results from those previous studies. A comparison of this study with other reported results is shown in Table 2 . The objective of this comparison was to highlight the similarities in performance metrics and effectiveness among previous studies. However, this study has advantage since several non COVID 19 case CXRs have been studied in here as well. As comparison, previous studies ( detection. The COVID . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted August 21, 2020. . Amid ongoing COVID 19 pandemic, a numerous routines include isolation and social distance seem to be temporary unpractical solution. While, COVID 19 vaccine is expected to take at least 18 months only if it works at all. Unfortunately, COVID 19 can mutate and become more aggressive. Learning from this aforementioned situations, then it is recommended to consider the use of AI and machine learning aided COVID 19 detection as one of the solution for the COVID 19 CXR analysis. Following COVID 19 cases, the demand for CXR will grow exponentially and resulted in the large number of CRX images waiting to be analyzed. Considering that the CRX image analysis may outnumber the capacity of all clinical settings, then applying AI and machine learning aided decision supports are considered to be very useful to support the frontline health practitioners. Correlation of Chest CT and RT-PCR Testing for Coronavirus Disease Artificial intelligence applied on chest Xray can aid in the diagnosis of COVID-19 infection: a first experience from Lombardy Can AI help in screening Viral and COVID-19 pneumonia? arXiv 2003.13145 Training and validating a deep convolutional neural network for computer-aided detection and classification of abnormalities on frontal chest radiographs Day Level Forecasting for Coronavirus Disease (COVID-19) Spread: Analysis, Modeling and Recommendations Time course of lung changes on chest CT during recovery from 2019 novel coronavirus (COVID-19) pneumonia Automatic X-ray COVID-19 Lung Image Classification System based on Multi-Level Thresholding and Support Vector Machine Clinical characteristics and intrauterine vertical transmission potential of . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity