The Role of Imaging in the Detection of Non-COVID-19 Pathologies during the Massive Screening of the First Pandemic Wave

During the COVID-19 pandemic induced by the SARS-CoV-2, numerous chest scans were carried out in order to establish the diagnosis, quantify the extension of lesions but also identify the occurrence of potential pulmonary embolisms. In this perspective, the performed chest scans provided a varied database for a retrospective analysis of non-COVID-19 chest pathologies discovered de novo. The fortuitous discovery of de novo non-COVID-19 lesions was generally not detected by the automated systems for COVID-19 pneumonia developed in parallel during the pandemic and was thus identified on chest CT by the radiologist. The objective is to use the study of the occurrence of non-COVID-19-related chest abnormalities (known and unknown) in a large cohort of patients having suffered from confirmed COVID-19 infection and statistically correlate the clinical data and the occurrence of these abnormalities in order to assess the potential of increased early detection of lesions/alterations. This study was performed on a group of 362 COVID-19-positive patients who were prescribed a CT scan in order to diagnose and predict COVID-19-associated lung disease. Statistical analysis using mean, standard deviation (SD) or median and interquartile range (IQR), logistic regression models and linear regression models were used for data analysis. Results were considered significant at the 5% critical level (p < 0.05). These de novo non-COVID-19 thoracic lesions detected on chest CT showed a significant prevalence in cardiovascular pathologies, with calcifying atheromatous anomalies approaching nearly 35.4% in patients over 65 years of age. The detection of non-COVID-19 pathologies was mostly already known, except for suspicious nodule, thyroid goiter and the ascending thoracic aortic aneurysm. The presence of vertebral compression or signs of pulmonary fibrosis has shown a significant impact on inpatient length of stay. The characteristics of the patients in this sample, both from a demographic and a tomodensitometric point of view on non-COVID-19 pathologies, influenced the length of hospital stay as well as the risk of intra-hospital death. This retrospective study showed that the potential importance of the detection of these non-COVID-19 lesions by the radiologist was essential in the management and the intra-hospital course of the patients.


Introduction
Since the end of December 2019, the COVID-19 pandemic has become a global health crisis [1][2][3]. Its diagnosis is most commonly based on RT-PCR [4,5]. In the early phase of the pandemic, healthcare had to deal with the inherent delay in diagnosis due to the processing time of the RT-PCR to confirm the diagnosis of suspicious cases. Facing a massive influx of suspected COVID-19 patients with limited availability of PCR diagnosis, clinicians took the opportunity of using chest imaging to help in the diagnostic approach. Indeed, patients infected with COVID-19 exhibit typical chest-CT lesions that can easily confirm the diagnostic suspicion.
The most frequent lesions on the positive COVID-19 chest CT showed good sensitivity for the diagnosis [6,7]. The main CT semiologies of COVID-19 pneumonia are bilateral ground-glass opacities located mainly in the subpleural and posterior areas. They are often associated with zones of focal condensation and crazy paving appearance in later forms. Therefore, CT scan has become a highly frequent test for screening and diagnosing patients having symptoms possibly inherent to COVID-19, such as dyspnea, polypnea and/or desaturation requiring hospital treatment (and also with a view to a preoperative assessment) [7,8]. Thus, thousands of chest CTs were performed at patients' admissions, whether with a confirmed or unconfirmed diagnosis of COVID-19 pneumonia. It must be noted that the diagnosis of COVID-19 on chest CT has a low specificity and mimickers may exist [9]. These various examinations have also indirectly led to de novo discoveries of non-COVID-19 associated chest abnormalities.
The outbreak of the COVID-19 pandemic also allowed the development of multiple artificial intelligence (AI) projects in order to assist the diagnosis, particularly by automating the assessment of the evolution of the disease and the quantification of lesions. Many tools were developed to classify and specify the lung abnormalities and quantify their extension (CO-RADS-for Reporting and Data System-and COVID-RADS, for example) [10,11]. At the University Hospital of Liège, we mainly used the COVIA system [12], which analyzes the whole lung field at a tomodensitometric level. Many other systems (such as the one developed by Robovision) were also used or tested. The COVIA system helps to identify an infection due to COVID-19 with 89.7% accuracy (95% CI: 84.0-93.9%). These analysis systems are essentially based on the identification of sensitive but not very specific COVID-19 lesions. On the other hand, the limitations of such automated or semi-automated approaches highlight the role of the radiologist's systematic diagnostic workup and his expertise in the detection and analysis of incidental occurrences of non-COVID-19 lesions.
In addition, the multiple chest CT performed in the context of screening for COVID-19 infection on a large population provides possibilities to retrospectively analyze the potential benefit of screening patients with chest CT. The value of screening for lung cancer by chest CT is already well established in patients with a history of smoking habits [13,14]. The benefits of annual screening for lung cancer by chest CT in a patient aged from 55 to 80 consuming at least 30 packs of cigarettes/year allow early diagnosis and better management [15]. The scanners performed as part of the COVID-19 screening provide a database for carrying out an analysis of unknown non-COVID-19 pathologies on a large population, which would thus reflect mass screening.
The aim of this article is to use the study of the occurrence of non-COVID-19-related chest abnormalities (de novo or known lesions) in a huge cohort of patients having suffered from confirmed COVID-19 infection and statistically correlate the clinical data and the occurrence of these abnormalities in order to assess the potential of increased early detection of lesions/alterations. Hypotheses are that a correlation exists between the clinical data, epidemiological data and these observed lesions, which would be more pronounced for de novo lesions. These correlations would then impact parameters such as length of stay or potential intra-hospital death.

Materials and Methods
We retrospectively studied a sample of adult patients (n = 362) admitted to the emergency department with proven COVID-19 infection between 2 March and 7 June 2020 and who benefited from one or more chest CT scans at the University Hospital of Liège.
The protocol was approved by the ethics committee of the University Hospital of Liège (Belgian number: B707201422832; ref:2020/127).
No written informed consent from all subjects (patients) is applicable as this study uses anonymous data.

Computerized Tomography Study
Images were retrospectively reviewed by P.C. (medicine resident with 4 years of experience in radiology). The pre-existing database has been selected with a PCR proving patient infection with COVID-19. We then quantified and described any identifiable intrathoracic lesions (see Table 1). Chest CTs were performed using a CT scanner (Siemens Edge Plus, GE Revolution CT, GE Brightspeed) available at the University Hospital of Liège [16,17]. Most of these were carried out according to a standard chest high-resolution computed tomography (HRCT) protocol with spiral volume acquisition in spontaneous contrast, thin sections and multiplanar reconstructions. The main scanning parameters were: tube voltage, 120 kVp; automatic tube current modulation; pitch, 0.99-1 mm; matrix, 512 × 512; slice thickness, 1 mm; and field of view, 31.6 cm. All images were then reconstructed with a slice thickness of 1.250-5 mm with the same increment. Other scans include angio-scans, low-dose thoracic scans and, to a lesser extent, thoraco-abdominal scans (the study of which is limited to the thorax and the first abdominal cuts with the vertebral body of L2 as a reference) in the thoracic scanner database. We also compared the newly identified lesions with any other images or written reports available in the Picture archiving and communication system (PACS), which is provided by the enterprise Imaging "AGFA". Pre-specified lesions which were already available in the medical imaging of the patient and the reports already available in PACS for this study were not considered de novo lesions.

Statistical Analysis
Continuous variables were described using mean and standard deviation (SD) or median and interquartile range (IQR) as appropriate. Qualitative variables were presented with frequency tables (numbers and percentages).
Logistic regression models were used to analyze the impact of patients' characteristics on the risk of presenting abnormalities on CT scans. Models were adjusted for age and gender. Results were presented using odd ratios (OR) and their 95% confidence intervals (95% CI) and "p" (p-value).
Linear regression models were used to analyze the impact of patients' characteristics on the length of hospital stay. Models were adjusted for age and gender, and lengths of stay were log-transformed. Results were presented as regression coefficient estimation, standard error (SE) and "p". Logistic regression models adjusted for age were used for risk of death at the hospital.
Results were considered significant at the 5% critical level (p < 0.05). Missing data were not replaced, and calculations were always performed on the maximum number of data available. Data analysis was carried out using SAS (version 9.4 for Windows). R (version 3.6.1) packages were used for the figures.

Results
We studied a group of 362 COVID-19-positive patients who performed a CT scan to diagnose and predict COVID-19-associated lung disease.
Patients' descriptions are listed in Table 2. The average age is 65 years with a male/female ratio of 1.3, and 85% of patients were non-smokers. Associated comorbidities were chronic renal failure (11%), diabetic patients (39%) and arterial hypertension (58%). Among patients, 31% suffered a cardiovascular disease, and 18% suffered a chronic lung disease. The proportion of oncological patients was 13%. The median length of stay was ten days for 91% of the sample hospitalized for COVID-19, and 22% of the cohort required a passage to intensive care, while 5.1% (17/330) faced intra-hospital death. In our cohort, we identified that 76% (280/362) of the population displayed non-COVID-19-related chest CT abnormalities. A total of 572 abnormalities were detected, with a proportion of 61% newly identified lesions, and 60.5% of the population was presenting new incidental lesions.
The retrospective analysis of incidental features (see Table 3) showed that calcifying atheromatosis was the most frequent incidental de novo abnormality (example of chest CT in Figure 1a), with 35% represented in the sample with a percentage of more than half (60%) with already known calcifying coronary atheromatosis. The other cardiovascular parameters were less represented, with aneurysmal dilation of the ascending thoracic aorta in 9.7% and the presence of a significant pericardial effusion in 3.9% (example of chest CT in Figure 1b). Lung diseases were found in 19.4% of the population: 18% of the cohort showed COPD-associated lesions (mainly emphysematous lesions) and 1.4% signs of pulmonary fibrosis. Thyroid goiter was detected in 22% of the cohort. Of note, 63% (49/78) of this abnormality were not previously identified based on the medical file review.
PACs-other images or written reports available in the Picture archiving and communication system (PACS); COPD-chronic obstructive pulmonary disease; SD-standard deviation. parameters were less represented, with aneurysmal dilation of the ascending thoracic aorta in 9.7% and the presence of a significant pericardial effusion in 3.9% (example of chest CT in Figure 1b). Lung diseases were found in 19.4% of the population: 18% of the cohort showed COPD-associated lesions (mainly emphysematous lesions) and 1.4% signs of pulmonary fibrosis. Thyroid goiter was detected in 22% of the cohort. Of note, 63% (49/78) of this abnormality were not previously identified based on the medical file review.
(a) (b)  Interestingly, we found that 8.8% of the patients were suffering from vertebral compression, whereas those lesions were not previously described in 3.3% of the total population.
The presence of a nodule on the chest CT scan, regardless of its origin, was found in 26% of the sample. Incidental nodules were then identified in 18.8% of the total cohort (example of chest CT in Figure 2b,c). A newly identified suspicious mass or a suspicious lesion greater than 3 cm was found in 5% (example of chest CT in Figure 2a).

Absent Present
Present Interestingly, we found that 8.8% of the patients were suffering from vertebral compression, whereas those lesions were not previously described in 3.3% of the total population.
The presence of a nodule on the chest CT scan, regardless of its origin, was found in 26% of the sample. Incidental nodules were then identified in 18.8% of the total cohort (example of chest CT in Figure 2b,c). A newly identified suspicious mass or a suspicious lesion greater than 3 cm was found in 5% (example of chest CT in Figure 2a).  The proportion of known and unknown lesions is balanced in the sample except for a more marked difference for suspicious nodule, thyroid goiter and the thoracic aortic aneurysm, with a known proportion of lesions at 2.5% known versus 7.2% of unknown aneurysmal lesion.

Discussion
In our study, we retrospectively identified that, among a cohort of 362 confirmed COVID-19 infected patients, incidental non-COVID-19-related chest lesions were mainly significant calcifying atheromatosis, suspicious nodule and thyroid goiter.
Age, male gender, having any COVID-19-associated comorbidity or active tobacco abuse were all specific risk factors for increased length of hospital stay (median (IQR) 10 days (6; 20)) or in-hospital death (5.1% of the hospitalized patients). Of interest, we showed that patients exhibiting at least one pre-existing CT abnormality were at increased risk of in-hospital death (risk multiplied by 2.8 (95% CI: (1.0; 7.9); "p" = 0.046)).
We identified that 60.5% of the population was presenting new incidental lesions. Of note, those were mainly calcified coronary disease, suspicious nodule and thyroid goiter. In our cohort, we found that 19% of patients were presenting incidental suspected lung nodule, which is in line with previous large screening cohort studies [27,28]. Based on the general recommendations of lung cancer screening, we also identified a subgroup of patients who were at higher risk of developing non-COVID-19 thoracic abnormalities. This population is characterized by older age, male gender and smoking status. In this context, a retrospective re-assessment of lung abnormalities after having performed a COVID-19 screening CT scan would be of interest and would increase the early detection of lesions/alterations. Cardiovascular and pulmonary diseases have become major causes of death worldwide. According to the World Health Organization, cardiovascular diseases account for a major part with 17.5 million deaths, followed by 8.2 million for oncological and pulmonary pathologies [29]. Interestingly, the most frequent incidental lesion in our study is calcifying coronary atheromatosis, which, therefore, can have a significant impact on the patient outcome. Our data are in line with previous studies, identifying a global prevalence of calcifying coronary atheromatosis approaching 40% in a population of 60-year-old women [30], ranging up to 75% in patients over 70 in this study. With respect to a previous study [31], risk factors associated with severe COVID-19 infection are known to be mainly cardiovascular conditions, in opposition to previous respiratory chronic lung diseases [32,33].
In our cohort, the probability of encountering a thoracic lesion increases with age. The likelihood of developing calcifying atheromatosis, COPD signs or suspicious thoracic mass increases with age. Patients with pre-existing respiratory or cardiac comorbidities did frequently benefit from previous thoracic explorations leading to dedicated systematic medical and radiological follow-up [34]. This could potentially explain why we found in our study more incidental lesions in younger patients because they are more likely to have fewer comorbidities and, therefore, less routine medical follow-up. Patients with smoking history exhibit a four times higher risk of having lung lesions, whereas chronic renal failure provides nine times increased risk.
Similar to previous studies [35,36], our data show that length of hospital stay (LOS) in COVID-19-positive patients is influenced by many comorbid factors such as renal failure and hypertension. Age and comorbidities were found to be strong predictors of hospital admission and, to a lesser extent, of critical illness and mortality in people with COVID-19 [37]. Incidental identification of non-COVID-19-related thoracic lesion also appears to be an intrinsic risk factor for increased hospital LOS. It underlies the impact on potential comorbidities in severe COVID-19 patients. We thus identified that pre-existing signs of pulmonary fibrosis and vertebral compression were associated with an increased length of hospital stay. It is therefore essential to be able to better identify those comorbid lesions in order to propose a holistic approach to patients who are at higher risk of experiencing severe COVID-19 disease. However, our study did not show that a non-COVID-19 thoracic abnormality influences intensive care unit LOS. Unsurprisingly, the impact of in-hospital death is also correlated with comorbidities.
The comorbidities association such as a smoking or oncological history has shown to provide an increased risk of in-hospital death. In the study of Dai and al., patients with cancer (notably hematologic cancer, lung cancer, or metastatic cancer (stage IV)) appear more vulnerable to SARS-CoV-2 [38]. In addition, in our study, the mortality is increased by 2.8 times in patients with at least one abnormality identified with a CT scan. The presence of an incidental abnormality alone increasing the risk of in-hospital death should be subjected to multivariate analysis. Recent studies on COVID-19 showed that hypercholesterolemia, diabetes or COPD were associated with worse clinical outcomes, including COVID-19-associated mortality [37][38][39][40].
Taking into consideration the need for rapid CT-scan evaluation in the burden of the COVID-19 pandemic, it is nevertheless important to carefully evaluate non-COVID-19related thoracic abnormalities as they are potentially influencing patient outcomes that can be overcome with artificial intelligence models.

Limitation
In this study, we join the data at the global level regarding the importance of certain non-COVID-19 pathologies in the general population. However, the selection bias inherent in the positive COVID-19 patients selected in the study must be considered. We have shown that cardiovascular pathologies such as calcifying coronary atheromatosis were one of the most frequent pathologies in patients over 62 years of age, as found in the data in the literature, but the real background of this study is the impact of these non-COVID-19 chest lesions on patient management. This study showed that these lesions could represent unrecognized comorbidities with a significant impact on the patient. They could be integrated into a better assessment of the risk profile, especially since these lesions are generally not detected by artificial intelligence systems in the detection of COVID-19 pneumonia. The systematic approach of the radiologist, here more particularly, in the detection of non-COVID-19 lesions, thus plays a major role. The usefulness of the AI-based model in medical imaging is therefore complementary with the radiologist analysis in order to increase sensitivity and specificity of the diagnosis performance of chest CT.

Conclusions
This retrospective study showed that in the context of the pandemic, clinicians and radiologists had the chance to identify non-COVID-19-associated chest CT abnormalities. The high number of CT scan analyses performed also provides an opportunity to reevaluate the occurrence of any thoracic abnormalities than can be incidentally identified in a general population. Correlations exist between clinical data, epidemiological data and these observed lesions, which can potentially be more pronounced for de novo lesions. These correlations could be confirmed if the data from this study are included in a metaanalysis which can be used to improve general patient care.