Improving color and shape repeatability of tongue images for diagnosis by using feedback gridlines

Introduction: In traditional East Asian medicine, the color and shape features of a tongue reflect the health of the body. The repeatability of computerized tongue diagnosis is highly dependent on the image acquisition process. In this study, a method for providing frontal and profile feedback gridlines for the tongue image acquisition process was proposed and the resulting color and shape repeatability was assessed. Methods: Eight subjects were recruited and 120 images were acquired. For each subject, 15 tongue images were acquired with three different gridline types, including no gridline (Group I), frontal gridlines (Group II), and frontal and profile gridlines (Group III) in a pseudo-randomized order. To evaluate repeatability, intra-class correlation coefficients (ICCs) were calculated from the mean values and histogram fitting parameters for Commission Internationale de l'Éclairage L*. a*. b*, profile angles, and tongue area ratios (TARs) of four edge regions, respectively. Results: The ICCs for all the color features in Group I were lower as compared with those in Group II and additionally, the ICCs for all the color features in Group III were larger compared with those in Group II. The ICCs of the shape features were lower than those of the color features. The ICCs of TARs in Group I were lower than those in Groups II and III. From the tip to the root of the tongue region, the ICCs of mean luminance tended to decrease in all the groups, but the intensity of the decline in Group III was lower than those in Group I and Group II. Conclusions: The proposed tongue image acquisition process with frontal and profile gridlines improved the repeatability of color and shape features and potentially improve diagnostic accuracy of computerized tongue diagnosis.