Histogram-based quantitative evaluation of endobronchial ultrasonography images of peripheral pulmonary lesion

Abstract

Background: Endobronchial ultrasonography using a guide sheath (EBUS-GS) is an increasingly common bronchoscopic technique, but currently, no methods have been established to quantitatively evaluate EBUS images of peripheral pulmonary lesions. Objectives: The purpose of this study was to evaluate whether histogram data collected from EBUS-GS images can contribute to the diagnosis of lung cancer. Methods: Histogram-based analyses focusing on the brightness of EBUS images were retrospectively conducted: 60 patients (38 lung cancer; 22 inflammatory diseases), with clear EBUS images were included. For each patient, a 400-pixel region of interest was selected, typically located at a 3-to 5-mm radius from the probe, from recorded EBUS images during bronchoscopy. Histogram height, width, height/width ratio, standard deviation, kurtosis and skewness were investigated as diagnostic indicators. Results: Median histogram height, width, height/width ratio and standard deviation were significantly different between lung cancer and benign lesions (all p < 0.01). With a cutoff value for standard deviation of 10.5, lung cancer could be diagnosed with an accuracy of 81.7%. Other characteristics investigated were inferior when compared to histogram standard deviation. Conclusions: Histogram standard deviation appears to be the most useful characteristic for diagnosing lung cancer using EBUS images.