Non-Invasive Procedure in Differential Diagnosis of Sarcoidosis and Tuberculosis Lymph Nodes:  Radiomic Model of 18F-FDG PET-CT.

BACKGROUND AND AIM: Clinical and pathological features of two granulomatous diseases tuberculosis (TB) and sarcoidosis lymphadenopathy share similar properties. 18-F FDG Positron-Emission Tomography-Computed Tomography (18F-FDG PET-CT) is performed to discriminate two diseases. Even biopsy and culture via Endobronchial Ultrasonography (EBUS) sometimes did not get definite diagnosis. Radiomics can defined as high-throughput mining of radiological images. We aimed to investigate the role of radiomic analysis of these 18F-FDG PET/CT images in discrimination of TB and sarcoidosis Methods: All patients with mediastinal LAP who underwent EBUS biopsy were screened for inclusion. Among these patients, patients who were diagnosed with TB or sarcoidosis by pathological and microbiological methods were included in the study. Radiomic model and clinicoradiomic models were formed AUC, sensitivity and specificity values of models obtained by logistic regression results were calculated.

RESULTS: 54 tuberculosis and 163 sarcoidosis lymph nodes were analyzed. Gender, GLCM_Correlation and GLCM_Energy features were found to be important prognostic factors in distinguishing between sarcoidosis and tuberculosis (p: 0.012, OR: 2.423 (1.215-4.830, 95% CI); p<0.001, OR: 5.400 (2.108-13.830, 95%) CI); p<0.001, OR: 3.335 (1.693-6.571, 95% CI; respectively). The p, AUC, sensitivity, and specificity values of the obtained clinicoradiomic model were calculated as <0.001, 0.762 (0.651-0.798, 95% CI), 59.5% and 81.5%, respectively.

CONCLUSIONS: The model created with radiomics methods and clinical features gave significant results in distinguishing tuberculosis and sarcoidosis. This is promising for radiomic models that could replace invasive methods. It is expected that radiomic models will be used more in daily life in the future.