In December 2019, a new human virus species, called coronavirus 2019, was identified in Wuhan, Hubei Province, China which causes a disease named “COVID-19”. The virus can rapidly spread and cause acute respiratory syndrome due to the presence of pneumonia. On March 11, 2020, the WHO declared the disease as a pandemic due to its high prevalence. Clinical signs of COVID-19 are fever (85%), cough (70%), and shortness of breath (43%) as well as gastrointestinal and abdominal symptoms. In some cases, it can even be asymptomatic. The overall mortality rate for COVID-19 is 2.3% [1, 2]. Real time-polymerase chain reaction (RT-PCR) is the standard method for diagnosing the COVID-19 with high specificity, but it has some limitations. This test has a relatively high rate of false-negative due to mistakes in sampling and low sample size; therefore, its sensitivity is low (about 59–71%). Moreover, PCR is a time-consuming test and the limitation in the number of PCR kits delay the testing process, which can lead to delayed diagnosis and early initiation of quarantine during the pandemic [2, 3]. Computed tomography (CT) scan with high sensitivity is another method for early detection of COVID-19 and can compensate the low sensitivity of PCR. CT scan sensitivity is high (about 88–98%). Moreover, CT is a feasible modality with a short scan time and long throughput time [3,4,5,6]. By combining CT with PCR, the COVID-19 can be diagnosed earlier.
After the prevalence of COVID-19, extensive studies of lung CT images have been conducted to differentiate COVID-19-induced pneumonia from other causes of pneumonia, which are based on the visual evaluation of images [2,3,4,5,6,7,8,9,10,11,12,13]. The CT scan for diagnosing COVID-19 can show unusual pneumonia, often peripheral, with the bilateral distribution in the lungs. The patchy ground-glass opacities (GGOs) and consolidations are the most common findings of chest CT scan for the COVID-19. The GGO pattern is often multifocal, bilateral, peripherally extended, and consolidated due to interlobular and broncho vascular septal thickening [11, 12]. In the early stages of the disease, GGO may appear as a single focal lesion, mostly in the lower lobe of the right lung [10, 14].
The most common morphological patterns of GGOs are patchy and rounded ones followed by triangular and linear ones. Some studies have been reported that an angular GGO with pleural thickening is a specific radiological sign of the COVID-19. Crazy paving is also common due to thickened interlobular septa with intralobular reticulations [3,4,5]. The prevalence of GGO and consolidation on the chest CT scan in COVID-19 patients are 88% and 32%, respectively. Moreover, bilateral lung involvement and peripheral opacities are observed in 87% and 76% of the cases [4].
The findings of CT scan can be changed as the disease progresses. In the early stages of COVID-19 infection, small patchy GGO with thick vascular lumens is more common, although, 56% of patients have normal CT images in the first two days after the onset of clinical symptoms (fever, dry cough, etc.). In cases of disease progression, multiple GGOs, and some severe cases, the patients may have diffuse lesions in both lungs that appear as a "white lung". When the patient recovers, the pulmonary GGOs disappear in the CT image, and the squamous pattern of GGO and parenchymal fibrosis are not observed [4, 10, 14].
The major limitation of CT scan in diagnosing COVID-19 and differentiating it from other types of pneumonia is its low specificity (about 34%) [4, 6]. So, it seems to be very important to find new methods in medical imaging such as texture analysis, mathematical techniques to extract significant features from an image at different gray levels, assessing the heterogeneity of lesions, and improving the early diagnosis [15].
Radiomics, as a completely non-invasive method for quantitative analysis of medical images, has recently received considerable attention. The radiomics method objectively characterizes multiple types of lesions using the advanced quantitative features called “radiomics features” of medical images. These features are divided into two general categories: semantic and agnostic features. Semantic features are used to describe morphologic characteristics of lesions such as shape, size, location, etc., while agnostic features (e.g. textural features) use innovative mathematical procedures in a high-throughput way that may fail to be perceived by the naked eye [16,17,18,19].
This study aimed to explore the association of textural radiomics features with COVID-19 pneumonia and non-COVID pneumonia on lung CT scans; that may provide a noninvasive means to profound recognition of the radiomics pattern of COVID-19 pneumonia as well as help to differentiate between COVID and non-COVID patients.