Differential diagnosis of tuberculosis and sarcoidosis related mediastinal lymhp nodes using PET-CT radiomic analysis

<bold>Introduction:</bold> Differential diagnosis of sarcoidosis (SC) and tuberculosis (TB) mediastinal lymphadenitis (ML) is challenging, even patients undergo invasive procedures to get diagnosis. We aimed to investigate whether radiomic analysis of PET-CT images are useful or not in term of distinction between SC and TB. <bold>Material-Method:</bold> TB and SC patients with ML, who were biopsied with endobronchial ultrasound and diagnosed with pathological and culture methods were included in the study. We used PET-CT texture-volumetric analysis in this study. <bold>Results:</bold> 217 lymph nodes which belongs to the 63 patients [18 TB (%28.6), 45 SC] were evaluated. There was statistically significant difference between SC and TB groups among five texture parameters (GLCM_Energy, GLCM_Correlation, GLCM_Entropy log10&2, GLRLM_RLNU, GLZLM_GLNU). When gender, GLCM_Correlation and GLCM_Energy features were analyzed with logistic regression, it was seen that all 3 parameters were found to be important prognostic factors in distinguishing between SC and TB (radiomic model AUC: 0.724, clinicoradiomic model AUC: 0.761) (Figure 1). <fig><object-id>erj;64/suppl_68/PA650/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig> <bold>Conclusion:</bold> Our study is the first to differentiate TB and SC in ML using PET-CT radiomic analysis. The model created with radiomics methods and clinical features gave significant results in distinguishing TB and SC. It is expected that radiomic models will be used more in daily life in the future.