Philippe Thuillier1,2, David Bourhis3,4, Ulrike Schick5, Zarrin Alavi6, Catherine Guezennec3,4, Philippe Robin3,4, Véronique Kerlan7,3, Pierre-Yve Salaun3,4, Ronan Abgral3,4. 1. Department of Endocrinology, University Hospital of Brest, Brest, France - philippe.thuillier@chu-brest.fr. 2. EA GETBO 3878, University Hospital of Brest, Brest, France - philippe.thuillier@chu-brest.fr. 3. EA GETBO 3878, University Hospital of Brest, Brest, France. 4. Department of Nuclear Medicine, University Hospital of Brest, Brest, France. 5. Department of Radiotherapy, University Hospital of Brest, Brest, France. 6. EA-3878, INSERM CIC-1412 Medical University Hospital of Brest, Brest, France. 7. Department of Endocrinology, University Hospital of Brest, Brest, France.
Abstract
BACKGROUND: PET Textural indices could have an add-on diagnostic value for diagnosis of malignancy in patients with FDG-avid adrenal lesions. METHODS: Consecutive patients referred for a FDG-PET/CT to our nuclear medicine department from June 2012 to June 2017 were retrospectively screened. Inclusion criteria were: patients with a FDG-avid adrenal lesion (uptake≥liver background); malignant/benign lesion confirmed histologically or with follow-up imaging examination. Pheochromocytomas were not included in the analysis. For each adrenal lesion, 5 quantitative PET parameters (SUV<inf>max</inf>, MTV, TLG, TLR<inf>max</inf> and TLRmean<inf>)</inf> were calculated. Thirty-seven textural indices were extracted using LIFEx software®. Diagnostic performance to determine malignancy was assessed with a ROC analysis. Parameters with a significantly AUC>0.5 were selected and groups of highly correlated (r>0.8) parameters were created. A scoring system combining PET and textural indices was examined. RESULTS: PET textural indices were calculated for 53 lesions (37 malignant, 16 benign). Three PET metabolic parameters (SUV<inf>max</inf>, TLR<inf>max</inf>, TLRmean) and 13 textural indices had an AUC>0.5. Seven groups of highly correlated parameters (r>0.8) were extracted. For PET parameters, SUV<inf>max</inf> had the best AUC (0.89 95% CI [0.79-0.98]; cut-off=7.0). For textural indices, ZLNU had the best AUC (0.87 95% CI [0.78-0.96]; cut-off=34.7) and specificity of 100%. Three scores combining the best four textural indices alone (Contrast<inf>GLCM</inf>, LRHGE, SZE and ZLNU) or with one PET parameters (SUV<inf>max</inf>, TLR<inf>max</inf>) were developed but did not increase the diagnostic performance (AUC≤0.89). ZLNU was the best parameter to distinguish primary adrenal cancer from adrenal metastases in malignant lesions (P<0.001). CONCLUSIONS: Our study highlighted excellent diagnostic performance of several PET textural indices comparable to that of PET metabolic parameters. However, our results did not find any additional diagnostic value of textural indices when combined with metabolic parameters.
BACKGROUND: PET Textural indices could have an add-on diagnostic value for diagnosis of malignancy in patients with FDG-avid adrenal lesions. METHODS: Consecutive patients referred for a FDG-PET/CT to our nuclear medicine department from June 2012 to June 2017 were retrospectively screened. Inclusion criteria were: patients with a FDG-avid adrenal lesion (uptake≥liver background); malignant/benign lesion confirmed histologically or with follow-up imaging examination. Pheochromocytomas were not included in the analysis. For each adrenal lesion, 5 quantitative PET parameters (SUV<inf>max</inf>, MTV, TLG, TLR<inf>max</inf> and TLRmean<inf>)</inf> were calculated. Thirty-seven textural indices were extracted using LIFEx software®. Diagnostic performance to determine malignancy was assessed with a ROC analysis. Parameters with a significantly AUC>0.5 were selected and groups of highly correlated (r>0.8) parameters were created. A scoring system combining PET and textural indices was examined. RESULTS: PET textural indices were calculated for 53 lesions (37 malignant, 16 benign). Three PET metabolic parameters (SUV<inf>max</inf>, TLR<inf>max</inf>, TLRmean) and 13 textural indices had an AUC>0.5. Seven groups of highly correlated parameters (r>0.8) were extracted. For PET parameters, SUV<inf>max</inf> had the best AUC (0.89 95% CI [0.79-0.98]; cut-off=7.0). For textural indices, ZLNU had the best AUC (0.87 95% CI [0.78-0.96]; cut-off=34.7) and specificity of 100%. Three scores combining the best four textural indices alone (Contrast<inf>GLCM</inf>, LRHGE, SZE and ZLNU) or with one PET parameters (SUV<inf>max</inf>, TLR<inf>max</inf>) were developed but did not increase the diagnostic performance (AUC≤0.89). ZLNU was the best parameter to distinguish primary adrenal cancer from adrenal metastases in malignant lesions (P<0.001). CONCLUSIONS: Our study highlighted excellent diagnostic performance of several PET textural indices comparable to that of PET metabolic parameters. However, our results did not find any additional diagnostic value of textural indices when combined with metabolic parameters.