key: cord-1006427-iaxo4j4b authors: Asadi, Fariba; Shahnazari, Razieh; Bhalla, Nikhil; Payam, Amir Farokh title: Clinical evaluation of SARS-CoV-2 Lung HRCT and RT-PCR Techniques: Towards risk factor based diagnosis of infectious diseases date: 2021-04-30 journal: Comput Struct Biotechnol J DOI: 10.1016/j.csbj.2021.04.058 sha: 4ac9b7c9d73caf986bfd89296fe88f6647379ad5 doc_id: 1006427 cord_uid: iaxo4j4b This study uses image analysis techniques for comparative analysis of the lung HRCT features and RT-PCR of 325 suspected patients to COVID-19 pneumonia. Our findings propose more caution in the interpretation of RT-PCR data, promoting, instead, also the quantification of age and sex-based risk factors using HRCT images. Statistical analysis of our methodology reveals a direct relation between intensity, skewness and kurtosis of the radiological features and the gender of patients. Moreover, we investigate the effect of the age of patients on the appearance of COVID-19 pneumonia in the HRCT images. We have also applied our methodology to investigate the effect of time on the severity of COVID-19 pneumonia within the lungs. Subsequently, we find a strong relationship between image analysis and the informed medical diagnosis asserted by the radiologists. Additionally, our results also indicate increase in the severity of lung infection in the first and second week after the onset of the SARS-CoV-2 symptoms. Thereafter, a gradual decrease in the lung damage is observed during the third week. The proposed image analysis methodology can be used as a simple complementary tool for infectious disease diagnostics as demonstrated in this study with an example of SARS-CoV-2 to provide better understanding of the disease for drug and vaccine development. In December 2019, Severe Acute Respiratory Syndrome of Coronavirus 2 (SARS-CoV-2) appeared in Wuhan, China and shortly spread in the whole world [1] . Fever, dry cough, dyspnea, chest tightness and shortness of breath are common symptoms of SARS-CoV-2 which can lead to severe injury in the lungs and spreading of the virus to other organs such as kidneys, heart, thyroid, and adipose tissue [2] [3] [4] [5] [6] . In particular, SARS-CoV-2 infect human respiratory epithelial cells through the interaction of viral protein and the Angiotensin Converting Enzyme 2 (ACE2) receptors abundantly present on the cells forming the epithelial layers of the human respiratory tract [4] , [6] [7] [8] [9] [10] [11] . Several factors such as age, sex and comorbidities lead to variation in complications associated with COVID-19 pneumonia, ranging from asymptomatic cases to patients with severely damaged lungs. [12] [13] [14] . The repercussions of these symptoms and associated risk factors can lead to health complications in long term and in many cases these symptoms can be fatal in short term, if left unaddressed. As such the development of intervention medicine (e.g. specific drug) is still in pre-mature stage, while vaccination may have cause unwanted changes in the body in long term [15] . Therefore, precise detection and continuous informed monitoring of infection is urgently required to be established, one possible route can be through the customization/improvement in analysis of techniques routinely used in medical industry. Current detection methods which are widely used for diagnosis of SARS-CoV-2 are Reverse Transcription Polymerase Chain Reaction (RT-PCR), chest X-ray, High Resolution Computed Tomography (HRCT) scans, and the detection of some common biomarkers in the blood [16] [17] [18] [19] [20] . However, recent studies reported that the positive rate of RT-PCR is between 30% to 60% [20] [21] [22] [23] . This suggests that at initial stage of the disease, many infected cases may not be detected due to the low sensitivity of RT-PCR. In parallel, recent clinical investigations have revealed the advantages of Chest-HRCT (hereafter referred as HRCT) to demonstrate typical radiological features in the patients suspected to COVID-19 pneumonia, consisting both positive and negative RT-PCR results [20] , [22] [23] [24] [25] [26] . However, due to the infancy of the research in the HRCT and RT-PCR analysis of COVID-19 pneumonia, in addition to the limitation of HRCT such as low specificity, need of high level skills (combination of clinical and image development expertise) in HRCT image analysis and low sensivtity of HRCT for early stage rapid disease detection [27] [28] [29] , new and simple image analysis are required to enhance the capaciaty of HRCT and RT-PCR, For instance, understanding the association between risk factors and the findings of the HRCT in combination with RT-PCR may lead to more specific detection than the current state of art in the use of HRCT and RT-PCR for disease detection. In this context, we report clinical results of HRCT and RT-PCR of 325 patients suspected with COVID-19 pneumonia and evaluate the advantages of HRCT in comparison to the RT-PCR test. Furthermore, based on the quantitative and statistical image analysis on the HRCT data, we found the relation between sex and age as main risk factors in the enhancement of the disease and features identified in the HRCT. In order to investigate our hypothesis, RT-PCR and HRCT was conducted in the patients and new image analysis was performed in the patient data. Starting from HRCT outcomes, RT-PCR results have been analyzed and associated to the assertions made using image analysis. The obtained results are discussed with respect to age and sex-based risk factors to quantify with the severity of damage in the lungs due to the COVID-19 pneumonia. The All HRCT were performed using a GE scanner (Optima CT540, UK). A low-dose institutional protocol was applied with the main scanning parameters as follows: tube voltage: 100kVp; tube current: 10-202 mAs; automatic exposure control; slice thickness = 3.75 mm. CT images were acquired at full inspiration with the patient in supine position, and without administration of intravenous contrast medium. Two radiologists (F.A. and R.S., with 6 and 7 years of experience in interpreting chest CT images, respectively) blinded to RT-PCR results reviewed all HRCT images and decided on positive or negative HRCT findings by consensus. The epidemiologic history and clinical symptoms (fever and/or dry cough) were available for both readers. The radiologists classified the HRCT scan (according to Radiological Society of North America (RSNA) classification) as, typical, indeterminate, atypical and negative for COVID-19 pneumonia. A description of main HRCT features and lesion distribution has been performed. Quantitative and statistical image analysis was performed using ImageJ software, a java-based image processing tool developed by National Institute of Health (NIH) The diagram of our case selection is given in figure 1 . In our patients selection we follow Fleischner Society criteria [30] . According to HRCT findings and RSNA classification [31] , General steps involved within the analysis are highlighted in figure 2 . The flowchart in this scheme shows operation protocol for the analysis of HRCT images and subsequent analysis using parameters of mean, skewness and kurtosis. It should be noted that the step 5 in the protocol will analyze complete image (including scale bar and text within standard HRCT image) and therefore specific HRCT area for measurement must be manually selected for analysis. In this study we focus on the 57 patients, with RT-PCR positive. Among these patients, crazy paving appearance were seen in 18 of them (patchy: 8 patients, continues: 10 patients), Ground Glass Opacities (GGO) is seen in 18 patients, peribronchiovascular involvement has been observed in one patient, consolidation is seen in 13 patients (with air space appearance in 5 of them) and halo sign is observed in 1 patient. In addition, mix appearance is observed in 4 patients, normal HRCT is reported for 1 and cardiomegaly without other involvement has been seen for 1 patient. Figure 3 shows examples of main CT features observed in the patients with positive RT-PCR test. We also observed parenchymal involvement in 55 patients using HRCT; in particular the unilateral lung parenchymal involvement was seen in 6 of them while bilateral lung Out of 57 patients, 11 patients passed away. In these 11 patients the continuous crazy paving appearance was seen for 6 of them, GGO appearance was observed for 3 of a b c d e them and the consolidation appearance was seen in one of them. The peribronchovascular was seen in one patient too. Ten patients had bilateral involvement and for 1 dead patient the unilateral involvement was observed and 5 of them had pleural effusion. Figure 4 and Tables 2 and 3 In order to compare the sensitivity of HRCT with RT-PCR and study the relation c < 0.05 was considered statistically significant, with 95% confidence interval. Figure 5a shows the statistical results of comparison between cases with positive HRCT and both positive and negative RT-PCR. As seen from results, there is no significant differences between positive and negative RT-PCR in the intensity, skewness and kurtosis calculated from HRCT images. Moreover, it is observed that RNA virus detection cannot completely reply on the RT-PCR as shown by HRCT. RT-PCR negative hits found to be 42.3% for 33 patients out of all 78 patients tested with positive HRCT. It can be explained by the lack of sensitivity, insufficient stability, and relatively long processing of RT-PCR test [20] , [32] . Other factors which can affect the accuracy of RT-PCR are specimen source (lower or upper respiratory tract), performance of kits and the sampling time window which is related to the time of disease development with which RT-PCR test has been conducted. In order to study the relation between severity of damages and appearances of COVID-19 pneumonia in the lungs with the gender and age as two main risk factors in SARS-CoV-2 disease [12] [13] [14] , we have performed another image analysis. The results are given in figures 5d to j. The HRCT mean (5a), skewness (5b) and kurtosis (5c) show higher intensity values in RT-PCR positive compared to RT-PCR false negative. As it is depicted, generally the mean intensity for males is higher than females independent of RT-PCR results (5d). Furthermore, for both positive and negative RT-PCR results, the mean intensity of males is higher than the females. This is in agreement with recent RNA-sequential and single cells analysis which has been performed on SARS-CoV-2 infected samples to describe higher susceptibility in male than female to COVID-19 pneumonia [4] , [33] . According to report of [4] , significant correlations between ACE2 expression levels with CD8+ T cell enrichment level (0.20