Investigation of the relationship of CO-RADS and CT patterns with laboratory parameters in COVID-19 patients and a new perspective on the total CT scoring system.

BACKGROUND: It is important to determine the correlation of the CO-RADS classification and computed tomography (CT) patterns of the lung with laboratory data. To investigate the relationship of CO-RADS categories and CT patterns with laboratory data in patients with a positive RT-PCR test. We also developed a structured total CT scoring system and investigated its correlation with the total CT scoring system. METHOD: The CT examinations of the patients were evaluated in terms of the CO-RADS classification, pattern groups and total CT score. Structured total CT score values were obtained by including the total CT score values and pattern values in a regression analysis. The CT data were compared according to the laboratory data. RESULTS: A total of 198 patients were evaluated. There were significant differences between the CO-RADS groups in terms of age, ICU transfer, oxygen saturation, creatinine, LDH, D-dimer, high-sensitivity cardiac troponin-T (hs-TnT), CRP, structured total CT score values, and total CT score values. A significant difference was also observed between the CT pattern groups and oxygen saturation, creatinine and CRP values. When the structured total CT score values and total CT score values were compared they were observed to be correlated. CONCLUSIONS: Creatinine can be considered as an important marker for the CO-RADS and pattern classifications in lung involvement. LDH can be considered as an important marker of parenchymal involvement, especially bilateral and diffuse involvement. The structured total CT scoring system is a new system that can be used as an alternative.

A Case of Intestinal Perforation Associated with Mesenteric Thrombosis Due to Post-COVID-19 Syndrome.

PURPOSE: We aimed to present a case who developed intestinal ischemia and associated perforation and abscess due to Superior Mesenteric Vein (SMV) thrombosis caused by post-COVID-19 syndrome and discuss the preoperative Computed Tomography (CT) imaging findings used in diagnosis. CASE PRESENTATION: A 58-year-old patient presented to our clinic with a complaint of acute abdominal pain. His CT examination revealed thrombosis in SMV, congestion in the mesenteric venous structures, contamination in the mesentery, and thickening and dilatation of the jejunal loops due to ischemia. The patient had a history of acute COVID-19 infection. He had typical COVID-19 pneumonia findings (peripheral ground-glass opacities in both lung parenchyma predominantly in the lower lobe) on the thorax CT at that time. He was followed up with anticoagulant therapy. During his follow-up, a thoracic and abdominal CT was performed due to recurrent acute abdominal findings. On thorax CT, there was a web-like filling defect consistent with pulmonary embolism, traction bronchiectasis consistent with late findings of COVID-19 pneumonia, and poorly circumscribed subpleural ground glass opacities. On abdominal CT, in addition to mesenteric ischemia findings, loss of wall integrity was observed in the jejunal loops due to perforation and collection areas containing air consistent with an abscess. He was treated with small bowel resection and abscess drainage. CONCLUSION: Patients with acute COVID-19 infection should be followed up for the early diagnosis of serious symptoms that may develop due to post-COVID-19 syndrome, and contrast-enhanced CT should be the imaging method of choice to detect possible mesenteric vascular thrombosis in patients with acute abdominal symptoms.

Correlation of Indeterminate Lesions of COVID-19 Pneumonia Detected on Computed Tomography with RT-PCR Results.

BACKGROUND: The typical findings of COVID-19 pneumonia include multilobar groundglass opacities and consolidation areas observed predominantly in the basal and peripheral parts of both lungs in computed tomography. OBJECTIVE: The current study aimed to correlate indeterminate lesions of COVID-19 pneumonia detected on computed tomography with the results of the reverse transcription-polymerase chain reaction (RT-PCR) test. METHODS: Patients with high-resolution computed tomography images that were reported to contain indeterminate lesions in terms of COVID-19 pneumonia were included retrospectively in the study. The lesions were categorized and the patterns were classified. The RT-PCR-positive and the RTPCR- negative patients were compared. P<0.05 was accepted as the statistical significance limit. RESULTS: The RT-PCR-positive patients exhibited a higher rate of peripheral lesions. Limited consolidation areas were not detected in the RT-PCR-positive patients. In the RT-PCR-negative patients, the rates of acinar nodules and the tree-in-bud pattern were significantly higher. The RTPCR- negative patients had higher nodular contour features and lesion coalescence. In the subgroup consisting of lesions with ground-glass opacities and/or ground-glass opacity around the nodule, the rate of nodular contour positivity was significantly higher in the RT-PCR- positive patients. CONCLUSION: COVID-19 pneumonia should be suspected when peripheral indeterminate lesions are detected. When indeterminate lesions, such as tree-in-bud pattern, acinar nodules and limited consolidation area are detected, alternative diagnoses should be considered first, even if there are ground glass opacities accompanying these lesions.

Contribution of artificial intelligence applications developed with the deep learning method to the diagnosis of COVID-19 pneumonia on computed tomography.

INTRODUCTION: Computed tomography (CT) is an auxiliary modality in the diagnosis of the novel Coronavirus (COVID-19) disease and can guide physicians in the presence of lung involvement. In this study, we aimed to investigate the contribution of deep learning to diagnosis in patients with typical COVID-19 pneumonia findings on CT. MATERIALS AND METHODS: This study retrospectively evaluated 690 lesions obtained from 35 patients diagnosed with COVID-19 pneumonia based on typical findings on non-contrast high-resolution CT (HRCT) in our hospital. The diagnoses of the patients were also confirmed by other necessary tests. HRCT images were assessed in the parenchymal window. In the images obtained, COVID-19 lesions were detected. For the deep Convolutional Neural Network (CNN) algorithm, the Confusion matrix was used based on a Tensorflow Framework in Python. RESULT: A total of 596 labeled lesions obtained from 224 sections of the images were used for the training of the algorithm, 89 labeled lesions from 27 sections were used in validation, and 67 labeled lesions from 25 images in testing. Fifty-six of the 67 lesions used in the testing stage were accurately detected by the algorithm while the remaining 11 were not recognized. There was no false positive. The Recall, Precision and F1 score values in the test group were 83.58, 1, and 91.06, respectively. CONCLUSIONS: We successfully detected the COVID-19 pneumonia lesions on CT images using the algorithms created with artificial intelligence. The integration of deep learning into the diagnostic stage in medicine is an important step for the diagnosis of diseases that can cause lung involvement in possible future pandemics.