AbstractBackground and objective
Recent data suggest that grey-scale textural analysis on endobronchial ultrasound (EBUS) imaging can differentiate benign from malignant lymphadenopathy. The objective of studies was to evaluate grey-scale textural analysis and examine its clinical utility.Methods
Images from 135 consecutive clinically indicated EBUS procedures were evaluated retrospectively using MATLAB software (MathWorks, Natick, MA, USA). Manual node mapping was performed to obtain a region of interest and grey-scale textural features (range of pixel values and entropy) were analysed. The initial analysis involved 94 subjects and receiver operating characteristic (ROC) curves were generated. The ROC thresholds were then applied on a second cohort (41 subjects) to validate the earlier findings.Results
A total of 371 images were evaluated. There was no difference in proportions of malignant disease (56% vs 53%,P= 0.66) in the prediction (group 1) and validation (group 2) sets. There was no difference in range of pixel values in group 1 but entropy was significantly higher in the malignant group (5.95 vs 5.77,P= 0.03). Higher entropy was seen in adenocarcinoma versus lymphoma (6.00 vs 5.50,P< 0.05). An ROC curve for entropy gave an area under the curve of 0.58 with 51% sensitivity and 71% specificity for entropy greater than 5.94 for malignancy. In group 2, the entropy threshold phenotyped only 47% of benign cases and 20% of malignant cases correctly.Conclusions
These findings suggest that use of EBUS grey-scale textural analysis for differentiation of malignant from benign lymphadenopathy may not be accurate. Further studies are required.SUMMARY AT A GLANCE
This large, single-centre UK study of EBUS-TBNA adds to the minimal literature on the utility of grey-scale textural analysis in risk stratification of EBUS-TBNA samples that are negative for malignancy. This study casts doubt on the utility of this technique as an ancillary tool and further studies are needed prior to application in practice.