Optical differentiation between malignant and benign lymphadenopathy by EBUS using grey scale texture analysis

Abstract

Hillebrand et al. published in 2015 2 contains some very important observation/conclusion that we will consider for future perspectives of studies in patients with COPD.As discussed in our study, one of the limitations was the use of medication prior to the assessments of autonomic function and physical fitness which may have influenced the results of these tests; however, our assessments were performed considering the 'real-life' condition of these patients with COPD. However, there are differences in methodology between the studies that may explain their negative results. MarceliEdey's group only studied one first-order and one second-order greyscale parameter. Our study looked at many parameters and found five significant factors to analyse lymph nodes in the validation set by calculating a combined likelihood ratio. Each factor was statistically discriminatory for malignant from benign nodes, allowing multiplication of results, strengthening our likelihood ratios. This is particularly relevant because the bronchoscope model in Edey's paper (BF-UC260FW) uses an updated ultrasound processor to ours with more image adjustment modes possible. The authors state that image acquisition was standardized to default settings; however, differences in image acquisition by this processor mean a broader initial assessment across the full range of greyscale parameters would have been potentially more revealing. We stated a GI3 setting and a thermal index for soft tissue as our settings as well as machine defaults. We acknowledged that results applied only to the equipment we used and that multicentre studies using that equipment would be needed.In the recent study digital images were saved to removable media and a single still image was captured avoiding artefact. In our study digital video recording of the whole procedure was reviewed frame by frame to allow selection of the image which was completely free of even the slightest motion or balloon artefact and which gave the node at its largest allowing for the small rotations of image which inevitably occur. This allows such things as large bronchial vessels to be easily excluded from the image. Despite the expertise in the procedure by Edey's group, using only a single image retrospectively may have impacted image quality in a percentage of images and hence greyscale results.Edey's group reports mean size for all lymph nodes based on diameter and states regions of interest (ROI) were 'greater than 32 pixels'. We assume this meant that ROIs were 'greater than 32 × 32 pixels', as in our dataset, the smallest lymph node analysed was 1000 pixels in size and the largest malignant node was >91 000 pixels in size. The exact size is important as it can impact on the amount of greyscale data available to be analysed within each lymph node. Indeed our benign nodes were significantly smaller (as is often the case clinically) which may have contributed to their measured homogeneity by a range of pathological mechanisms. We agree that larger data sets could help to clarify this rel...

The optimal sequence for bronchial brushing and forceps biopsy in lung cancer diagnosis: a random control study

“…One reason for this variability is limitations in tissue sampling techniques, which can make it impossible to obtain the most representative area of neoplastic tissue. Numerous techniques have been devised to circumvent these limitations and thereby increase diagnostic yield; they range from the relatively simple and inexpensive, such as biopsy, washing, brushing, and conventional transbronchial needle aspiration (TBNA), to the sophisticated and expensive, including autofluorescence bronchoscopy, bronchoscopic cryotechnology, and endobronchial ultrasound (EBUS) (1,(6)(7)(8)(9). Concurrent application of different sampling techniques can improve the yield, such as the combination of forceps biopsy and brushing, as well as a combination of histology and cytology.…”

“…Autofluorescence imaging (AFI) video bronchoscopy has shown good diagnostic sensitivity, albeit poor specificity, for diseases of bronchial mucosa (8). EBUS combined with TBNA and EBUS-guided transbronchial biopsy (TBB) are highly specific and sensitive methods for examining submucosal lesions, mediastinal lymph nodes and peripheral pulmonary lesions (9,13,14). All these approaches can improve the ability of bronchoscopy to diagnose lung cancer, but most require expensive instrumentation out of reach for primary hospitals around the world and for many health care centers in developing countries.…”