RESUMEN
OBJECTIVE@#To investigate a new noninvasive diagnostic model for nonalcoholic fatty liver disease (NAFLD) based on features of tongue images.@*METHODS@#Healthy controls and volunteers confirmed to have NAFLD by liver ultrasound were recruited from China-Japan Friendship Hospital between September 2018 and May 2019, then the anthropometric indexes and sampled tongue images were measured. The tongue images were labeled by features, based on a brief protocol, without knowing any other clinical data, after a series of corrections and data cleaning. The algorithm was trained on images using labels and several anthropometric indexes for inputs, utilizing machine learning technology. Finally, a logistic regression algorithm and a decision tree model were constructed as 2 diagnostic models for NAFLD.@*RESULTS@#A total of 720 subjects were enrolled in this study, including 432 patients with NAFLD and 288 healthy volunteers. Of them, 482 were randomly allocated into the training set and 238 into the validation set. The diagnostic model based on logistic regression exhibited excellent performance: in validation set, it achieved an accuracy of 86.98%, sensitivity of 91.43%, and specificity of 80.61%; with an area under the curve (AUC) of 0.93 [95% confidence interval (CI) 0.68-0.98]. The decision tree model achieved an accuracy of 81.09%, sensitivity of 91.43%, and specificity of 66.33%; with an AUC of 0.89 (95% CI 0.66-0.92) in validation set.@*CONCLUSIONS@#The features of tongue images were associated with NAFLD. Both the 2 diagnostic models, which would be convenient, noninvasive, lightweight, rapid, and inexpensive technical references for early screening, can accurately distinguish NAFLD and are worth further study.
Asunto(s)
Humanos , Enfermedad del Hígado Graso no Alcohólico/diagnóstico por imagen , Ultrasonografía , Antropometría , Algoritmos , ChinaRESUMEN
ObjectiveTo evaluate the effectiveness and consistency of three commonly used early colorectal cancer screening models for advanced colorectal adenoma as a noninvasive means， and to assess the predictive value of traditional Chinese medicine （TCM） tongue images in the models. MethodsPatients diagnosed with colorectal adenoma who underwent colonoscopy and pathological examination were selected as the study participants. Basic clinical data and tongue image were collected. The prediction models of Asia-Pacific colorectal screening （APCS） model， its revision （M-APCS） and colorectal neoplasia predict （CNP） model were applied to compare the predictive effects of the three models on advanced stage adenomas of the colon， the differences in clinical data and traditional Chinese medicine tongue characteristics among patients with different degrees of adenomas， and the similarities and differences in tongue characteristics among the models. The discriminative ability of the three risk models was evaluated using the area under the curve （AUC） and receiver operating characteristic （ROC） curves. The calibration was assessed using the Kuder-Richardson coefficient and the Hosmer-Lemeshow test for consistency analysis. ResultsA total of 227 patients with adenoma were analyzed， including 104 patients （45.82%） with advanced adenoma. In the detection of advanced adenoma， those with greasy coating （70 cases， 67.3%） were higher than those without greasy coating （34 cases， 32.7%， P＜0.05）. After multivariate analysis， the odds ratio （OR） value of non-greasy coating was 0.371 （0.204~0.673， P＜0.01）， indicating that non-greasy coating was a protective factor for advanced adenomas. Among the three risk models， the detection rate of advanced adenoma in the high-risk group with APCS was the highest （63.3%）， which was 1.49 times and 2.04 times that of the medium-risk group （42.6%） and the low-risk group （31.1%， P＜0.01）. The detection rate of advanced adenomas in high-risk groups of M-APCS and CNP was slightly higher than that in moderate or low risk groups （P＞0.05）. The proportion of yellow and greasy coating in high-risk group was higher than that in the medium-risk or low-risk group （P＜0.05）. For the ability to distinguish advanced and non-advanced adenomas， the AUC of APCS was 0.629 （95% CI： 0.556~0.702） and was higher than that of M-APCS （0.591） and CNP （0.586）. In calibration evaluation， Cronbach's alpha was 0.919 （＞0.7）， which indicated that the three models were consistent. In the correlation matrix， the correlation coefficients between APCS model and M-APCS model， and CNP model were 0.794 and 0.717， respectively， and the correlation coefficients between M-APCS model and CNP model were 0.873， Hosmer-Lemeshow χ2 =2.552， P＞0.05， which suggested that the three models had good calibration ability. ConclusionAll three models demonstrate the efficiency to identify advanced colorectal adenoma， and their calibration ability is considered to be good. Among the three models， the APCS exhibits the highest recognition efficiency， however， the recognition accuracy of the APCS model needs improvement. The presence of a greasy coating is identified as one of the potential predictors of advanced adenoma. Consequently， it can be considered for inclusion in the risk model of advanced colorectal adenoma to enhance the accuracy.
RESUMEN
Tongue diagnosis in Kampo medicine is considered to be a diagnostic method that can provide information about a patient's constitution and medical condition. We have identified the following problems in tongue diagnosis: the influence of external environmental factors such as light source, room temperature, and dryness, as well as subjective factors that depend on the knowledge and experience of medical doctors. To overcome these problems and to support Kampo diagnosis, we developed the Tongue Image Analyzing System (TIAS). Regarding color, objective numerical values L*a*b* were measured and standardization and objectification were promoted. We introduce some of the progress that has been made over the past 15 years since the development of TIAS.
RESUMEN
This study was aimed to observe the tongue image characteristics at different renal function stages among 157 chronic kidney disease (CKD) patients. Tongue images of 157 CKD cases and 30 healthy people were recorded with the instrument designed by Shanghai University of Traditional Chinese Medicine. Parameters of the tongue image were compared among different renal function stages. The results showed that compared with the control group, there were obvious decreasing on the R and L indexes in groups of different renal function stages (P < 0.05). Compared with the control group, there were obvious decreasing on the G and B indexes in groups of CKD1, CKD2, CKD3 and CKD4 stages (P < 0.05). Compared to CKD4 stage, there were obvious decreasing on the R, G, B, L indexes from group of CKD1 and CKD2 stage (P< 0.05). Compared with the group of CKD5 stage, there were obvious decreasing in the group of CKD1, CKD2, and CKD3 (P< 0.05). Compared with the control group, the cracking index was obvi-ous increased in 5 stages (P < 0.05). There was obvious difference in the greasy index between the group of CKD5 stage and the control group (P< 0.05). Compared with the control group, there was obvious difference in the peeling index from the group of CKD3, CKD 4, CKD5 stage (P < 0.05). Along the decreasing of renal function, compared with the group of CKD5 stage, there was obvious decreasing on the greasy index and peeling index in the group of CKD1 and CKD2 (P< 0.05). It was concluded that Chinese medicine tongue and face diagnosis instrument can pro-vide certain objective evidence for the clinical diagnosis of CKD.
RESUMEN
Objective :To investigate relationship between the glossoscopy objectivised image and kidey-biopsy pathology type of patients with primary glomerulopathy.Methods: The tongue image(including color clustering of tongue images and tongue shapes) of 66 patients with primary glomerulopathy were taken by Snakes Algorithm of Tongue Image Handling System and pathology diagnosis of kidney was also done in the same time,t-test was adopted to analyse the correlation of the glossoscopy objectivised image and kidey-biopsy pathology type of patients.Result: The result show that there are significant differences among tongue images of different pathology types(P
RESUMEN
Objective To extract the color feature of tongue image.Methods The tongue image was divided into 8?8 dynamic rectangular grids. Then the color histogram of each grid was calculated. All of them composed a high-dimensional vector (we call it the original feature vector).Finally, the dimensionality of the original feature vector was reduced by using Landmark Isomap. Results Experiments on 3 594 tongue images showed that several color features of tongue image, such as the color of tongue substance, the color of tongue coat, the shades of tongue color, and so on could be extracted. Conclusion This proposed method can be used to extract the color feature of tongue image effectively.
RESUMEN
Objective To develop a kind of tongue image analysis instrument which is used to deal with the deviation problem on the process of tongue image analysis currently. Methods The system included two parts:hardware and software. The hardware was made up of special tongue image acquisition device and computer. With the application of digital image analysis technology, the instrument implements functions such as tongue image collection, image preprocess, image segmentation and feature analysis. Results After a certain amount of sample study process, the accuracy of tongue shape recognition was about 90%. Conclusion This instrument can achieve the aim of tongue information objectification and are widely used for clinical medicine and teaching.
RESUMEN
Based on the structure of TCM tongue image recognition system,digital image processing technology in tongue objectiveness research is analyzed,especially on the methods and principles of color correction,image segmentation,pattern recognition,etc.The future development is also discussed.