1.上海中医药大学附属曙光医院,上海市中医药研究院肝病研究所,肝肾疾病病证教育部重点实验室,上海市中医临床重点实验室(上海 201203)
2.上海中医药大学交叉科学研究院(上海 201203)
3.江西中科九峰智慧医疗科技有限公司(江西 南昌 330095)
俞晓菡,女,硕士研究生,主要从事慢性肝病中医证候客观化研究工作
张华,研究员,硕士研究生导师;E-mail:lnutcmzh@126.com
扫 描 看 全 文
俞晓菡,余上海,惠梦雨,等.湿热内蕴证、肝郁脾虚证非酒精性脂肪性肝病患者眼底图像特征分析[J].上海中医药杂志,2023,57(2):94-100.
YU Xiaohan,YU Shanghai,HUI Mengyu,et al.Analysis of fundus image features in NAFLD patients with internal accumulation of damp⁃heat syndrome and liver depression and spleen deficiency syndrome[J].Shanghai Journal of Traditional Chinese Medicine,2023,57(2):94-100.
俞晓菡,余上海,惠梦雨,等.湿热内蕴证、肝郁脾虚证非酒精性脂肪性肝病患者眼底图像特征分析[J].上海中医药杂志,2023,57(2):94-100. DOI: 10.16305/j.1007-1334.2023.2207051.
YU Xiaohan,YU Shanghai,HUI Mengyu,et al.Analysis of fundus image features in NAFLD patients with internal accumulation of damp⁃heat syndrome and liver depression and spleen deficiency syndrome[J].Shanghai Journal of Traditional Chinese Medicine,2023,57(2):94-100. DOI: 10.16305/j.1007-1334.2023.2207051.
目的,2,运用机器学习方法分析非酒精性脂肪性肝病(NAFLD)常见中医证型(湿热内蕴证、肝郁脾虚证)患者的眼底血管图像变化特征,以期为中医目诊及肝病证候分类提供客观依据。,方法,2,借助眼底照相机采集NAFLD患者、健康人群的眼底血管图像,同时收集性别、年龄等一般资料,以及临床症状信息,通过改进的U-Net模型分割眼底血管,采用计算机视觉技术提取血管颜色、形态及Haralick纹理特征,并以健康人群为对照运用决策树模型建立疾病预测模型。运用SPSS 26.0及R语言分析湿热内蕴、肝郁脾虚型NAFLD患者的眼底图像特征,以及眼底图像特征与一般资料、临床症状等影响因素间的相关性。,结果,2,①共纳入NAFLD患者126例,其中湿热内蕴证61例、肝郁脾虚证44例、无典型证21例,分别设为湿热内蕴组、肝郁脾虚组、无典型证组;纳入健康人110例,设为健康组。②基于决策树模型建立NAFLD疾病诊断模型,准确率为0.972,灵敏度为0.970,特异性为0.975,AUC为0.995,可有效用于对NAFLD患者眼底血管特征的分析。③在血管形态和颜色特征方面,与无典型证组比较,湿热内蕴及肝郁脾虚组血管总面积降低,细小分支比例、弯曲程度、细小分支弯曲程度升高(,P,<,0.05)。在血管纹理特征方面,与无典型证组相比,湿热内蕴组角二阶矩升高、差分熵降低(,P,<,0.05),肝郁脾虚组对比度、差分方差、差分熵降低,逆差距升高(,P,<,0.05)。④共筛选得到NAFLD的6个常见症状,分别为乏力、口干、夜间睡眠不足、大便稀薄不成形、眼睛干涩及白天困倦欲睡。⑤相关性分析结果显示,NAFLD患者年龄与特征值角二阶矩、血管弯曲程度及逆差距呈正相关(,P,<,0.05),与血管总面积、平均宽度、颜色强度方差、差分熵、相关性、熵、和熵及信息度量f13呈负相关(,P,<,0.05);性别与血管细小分支弯曲程度呈负相关(,P,<,0.05);眼睛干涩与血管总面积、平均宽度及信息度量f13呈负相关(,P,<,0.05);大便稀薄不成形与血管弯曲程度、对比度及差分方差呈正相关(,P,<,0.05),与逆差距呈负相关(,P,<,0.05)。,结论,2,湿热内蕴证与肝郁脾虚证NAFLD患者眼底图像血管纹理存在特异性改变,可为基于眼底血管特征的NAFLD中医证候分类研究提供数据支撑。
Objective,2,To provide an objective basis for visual diagnosis theory of traditional Chinese medicine (TCM) and classification of liver disease syndromes and symptoms by analyzing the changes of fundus vascular images with the machine learning method in patients with non-alcoholic fatty liver disease (NAFLD) of common TCM syndromes (internal accumulation of damp-heat syndrome, liver depression and spleen deficiency syndrome).,Methods,2,General data such as gender and age, as well as clinical symptoms were collected from NAFLD patients and healthy controls, and their fundus vascular images were collected with professional fundus cameras. The fundus vessels were segmented by the improved U-Net model, and the color, morphology and Haralick texture features were extracted by computer vision technology. A decision tree model was used to develop a disease prediction model with healthy people as controls. The fundus image features of NAFLD patients with internal accumulation of damp-heat syndrome and liver depression spleen deficiency syndrome, as well as the correlation between the fundus image features and influencing factors such as general data, clinical symptoms and so on were analyzed by SPSS 26.0 and R language.,Results,2,①A total of 126 NAFLD patients with different syndromes were included (61 cases with internal accumulation of damp-heat syndrome, 44 cases with liver depression and spleen deficiency syndrome, and 21 cases with nontypical syndrome), and were divided into different groups accordingly. The healthy group included 110 healthy individuals as controls. ②The diagnosis model of NAFLD disease was established based on the decision tree model, with an accuracy of 0.972, a sensitivity of 0.970, a specificity of 0.975, and an AUC of 0.995, which can be effectively used for the analysis of fundus vascular features of NAFLD patients. ③In terms of vessel morphology and color, the internal accumulation of damp-heat syndrome group and liver depression and spleen deficiency syndrome group had smaller area, and higher small rate, curved and small curved than the nontypical syndrome group (,P,<,0.05). In terms of vascular texture, the internal accumulation of damp-heat syndrome group had higher value of asm and lower value of den than the nontypical syndrome group (,P,<,0.05), and the liver depression and spleen deficiency syndrome group had lower values of dva and den, and higher value of idm than the nontypical syndrome group (,P,<,0.05). ④A total of 6 common symptoms of NAFLD were obtained by screening, which were fatigue, dry mouth, lack of sleep at night, thin and unformed stools, dry eyes and sleepiness and desire to sleep during the day. ⑤The results of correlation analysis showed that the age of NAFLD patients was positively correlated with asm, curved and idm (,P,<,0.05), and negatively correlated with area, width, bsd, den, cor, ent, sen and f13 (,P,<,0.05), and gender was negatively correlated with small curved (,P,<,0.05). Dry eyes were negatively correlated with area, width and f13 (,P,<,0.05); thin and unformed stools were positively correlated with curved, con and dva (,P,<,0.05), and negatively correlated with idm (,P,<,0.05).,Conclusion,2,The presence of specific changes in the vascular texture of fundus images of NAFLD patients with internal accumulation of damp-heat syndrome and liver depression and spleen deficiency syndrome may provide data support for the study of the TCM syndrome and symptom classification of NAFLD based on fundus vascular features.
非酒精性脂肪性肝病中医证候中医诊断机器学习计算机视觉眼底血管图像
non-alcoholic fatty liver diseasetraditional Chinese medicine syndrome and symptomTCM diagnosismachine learningcomputer visual technologyfundus vascular image
俞洋. 眼的胚胎发育与五轮学说[J]. 辽宁中医杂志,2020, 47(10): 51-53.
MAJI D, SEKH A A. Automatic grading of retinal blood vessel in deep retinal image diagnosis[J]. J Med Syst, 2020, 44(10): 180.
MUNDI M S, VELAPATI S, PATEL J, et al. Evolution of NAFLD and its management[J]. Nutr Clin Pract, 2020, 35(1): 72-84.
ZHOU J, ZHOU F, WANG W, et al. Epidemiological features of NAFLD from 1999 to 2018 in China[J]. Hepatology, 2020, 71(5): 1851-1864.
覃映霖,胡振斌,陈彩凤,等. 非酒精性脂肪肝证候分布规律探析[J]. 中国中医药现代远程教育,2021, 19(17): 45-48.
王楠,王阳,袁慧琴,等. 脂肪肝严重程度与中医证型关系的Meta分析[J]. 世界中医药,2020, 15(5): 748-754.
田园硕,陈瑞琳,周春梅,等. 不同中医证型非酒精性脂肪性肝炎患者肝硬度值及影响因素分析[J]. 中西医结合肝病杂志,2022, 32(1): 35-37.
顾立梅,曹培让,顾超,等. 非酒精性脂肪肝中医证型及临床生化指标相关性研究[J]. 南京中医药大学学报,2019, 35(6): 738-740.
中华医学会内分泌学分会. 非酒精性脂肪性肝病与相关代谢紊乱诊疗共识(第二版)[J]. 临床肝胆病杂志,2018, 34(10): 2103-2108.
国家药品监督管理局. 中药新药临床研究指导原则(试行)[M]. 北京:中国医药科技出版社,2002.
胡鑫才,张华,周扬,等. 乙肝后肝硬化患者报告结局评价量表条目的建立及筛选[J]. 中华中医药杂志,2021, 27(6): 1526-1530.
XIAN Y X, WENG J P, XU F. MAFLD vs. NAFLD: shared features and potential changes in epidemiology, pathophysiology, diagnosis, and pharmacotherapy[J]. Chin Med J, 2020, 134(1): 8-19.
黄超原,李炜,程轶敏,等. 基于中西医病证特点的非酒精性脂肪性肝病模型分析[J/OL]. 中药药理与临床,2021[2022-07-14]. https://kns.cnki.net/kcms/detail/51.1188.R.20211206.1409.018.htmlhttps://kns.cnki.net/kcms/detail/51.1188.R.20211206.1409.018.html.
李茜,纪龙珊,张鑫,等. 运用化湿活血法治疗代谢相关脂肪性肝病经验[J]. 中华中医药杂志,2022, 37(3): 1473-1476.
蔡虹,曹永龙,张如棉,等. 康良石基于“浊脂”理论辨治脂肪肝经验撷英[J]. 上海中医药杂志,2021, 55(7): 25-28.
李翔,夏飞,邓颖,等. 大数据时代下的中医现代目诊数字化平台建设[J]. 中医学报,2020, 35(1): 19-22.
朱会明,贾微,刘平,等. 传统目诊理论及技术的现代应用与发展[J]. 中华中医药杂志,2019, 33(2): 635-637.
XIAO W, HUANG X, WANG J H, et al. Screening and identifying hepatobiliary diseases through deep learning using ocular images: a prospective, multicentre study[J]. Lancet Digit Health, 2021, 3(2): e88-e97.
YUAN T H, YUE Z S, ZHANG G H, et al. Beyond the liver: liver-eye communication in clinical and experimental aspects[J]. Front Mol Biosci, 2021, 8: 823277.
高悦,余上海,赵超群,等. 123例肝硬化患者眼底血管变化观察与分析[J]. 中国中医眼科杂志,2021, 31(8): 553-560.
缪文茹. 益气健脾消脂汤治疗新加坡非酒精性脂肪肝的临床研究[D]. 南京:南京中医药大学,2021.
KASPER P, MARTIN A, LANG S, et al. NAFLD and cardiovascular diseases: a clinical review[J]. Clin Res Cardiol, 2021, 110(7): 921-937.
JAYACHANDRAN A, DAVID D S. Textures and intensity histogram based retinal image classification system using hybrid colour structure descriptor[J]. Biomed Pharmacol J, 2018, 11(1): 577-582.
HOUSE M J, BANGMA S J, THOMAS M, et al. Texture-based classification of liver fibrosis using MRI[J]. J Magn Reson Imaging, 2015, 41(2): 322-328.
曾明明,殷鑫. 非酒精性脂肪性肝病中医证候分布规律初探[J]. 山西中医学院学报,2016, 17(2): 41-42.
0
浏览量
90
下载量
0
CSCD
0
CNKI被引量
关联资源
相关文章
相关作者
相关机构