Premium
Validation of the laryngopharyngeal reflux color and texture recognition compared to pH‐probe monitoring
Author(s) -
Du Chen,
ALRamahi Jehad,
Liu Qingsong,
Yan Yan,
Jiang Jack
Publication year - 2017
Publication title -
the laryngoscope
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.181
H-Index - 148
eISSN - 1531-4995
pISSN - 0023-852X
DOI - 10.1002/lary.26182
Subject(s) - hue , receiver operating characteristic , laryngopharyngeal reflux , artificial intelligence , vocal folds , texture (cosmology) , pattern recognition (psychology) , medicine , reflux , computer science , pathology , larynx , surgery , disease , image (mathematics)
Objective/Hypothesis The objective of this study was to determine the validity of our laryngopharyngeal reflux (LPR) diagnostic system from our previous study (Witt et al.[14][Witt DR, 2014]) against the results of a standard pH probe monitoring. We hypothesized that subjects with abnormal pH probe results demonstrate color and texture abnormalities that would be classified as LPR according to artificial neural network (ANN) analysis. Study Design Retrospective analysis. Methods Eighty‐two subjects, including 18 pH‐positive, 11 pH‐negative, and 53 control subjects were tested for LPR through multichannel intraluminal impedance 24‐hour pH (MII‐24pH) monitoring. Laryngoscopic images of all subjects were obtained. The hue and texture values of seven areas of interest, including true vocal folds, false vocal folds, arytenoids, and interarytenoid, were quantified using a hue calculation and two‐dimensional Gabor filtering. These served as inputs for the ANN. This was used to classify images through pattern recognition, and a receiver operating characteristic (ROC) analysis was performed to determine the effectiveness of the diagnosis. Results Classification accuracy for the combined hue and texture was 87.40%, with an area under the ROC curve of 0.910. Conclusion Although a previous study conducted classification based on RFS, this study suggests that color and texture analysis may be used to classify images based on the results of pH probing, a more objective approach for diagnosis. Additional studies should include more subjects to produce an even more accurate reading, and will use the color/texture analysis tool to test and confirm this application in a clinical setting. Level of Evidence 3B. Laryngoscope , 127:665–670, 2017