基于多視圖舌象特征融合的中醫(yī)證型辨識(shí)
關(guān)鍵詞:舌診;證型辨識(shí);特征融合;多視圖
中圖分類號(hào):TP183 文獻(xiàn)標(biāo)志碼:A 文章編號(hào):1001-3695(2025)07-026-2116-07
doi:10.19734/j. issn.1001-3695.2024.11.0471
Abstract:Fortongue diagnosis inTCM(traditional Chinese medicine),clinical physiciansobservethequality,coating,and baseof the tongue to understand the patient’s health status and pathologicalchanges.Existing tongue diagnosis auxiliary syndromediferentiation modelslack comprehensiveanalysisand ignore thecomplementarycharacteristicsofthe tonguesurface and tongue base.To solvethe problem,thispaper proposedasyndrome identification model basedon multi-view tongue featurefusion(SI-MTF).Bycolecting3348tonguesurfaceandtonguebaseimagesandrequestingtraditional Chinese medicine physicianstolabeltheimages with syndrome types,thispaperconstructedatongueimagedatasetand proposedSI-MTF.Firstly,this methodextractedthetongue surfaceand tongue base regions basedon Mask R-CNNand NanoDet,andthenconstructedaconvolutional neural network with denseconnectionsto extractandfusetonguesurfaceandtonguebase features.SI-MTF extractedcolor,shape,andtexturefeaturesof the tongue image throughchannelmeanof HSVcolormodel,localbinarypatternalgorithm,andGaborfilter.Duringthetrainingstage,thismethodusedateacher-studentnetwork toimprovetheability offusingtonguesurfaceandtonguebase.Finall,itusedfullyconnectedlayersandsoftmaxfunctiontolearntherelationship between fusion features and syndrome types for achieving TCM syndrome identification.Basedonthe tongue image dataset, this paper conductedcomparative experiment,parameter discussion,ablation experiment,and robustness experiment,and then the method obtained an accuracy of 76.75% ,improved by 7.51 percentage points compared to the baseline method. Theexperimentalresultsshowthat thecomprehensiveanalysisofthetonguesurfaceandtonguebasebasedonmultiviewcaneffectively improve theperformanceof the tongue diagnosis auxiliary syndrome diferentiation model.
Keywords:tongue diagnosis;syndrome identification;feature fusion;multi-view
0 引言
望診是中醫(yī)四診之首,其中舌診作為望診中的關(guān)鍵流程,醫(yī)師通過觀察舌質(zhì)、舌苔和舌底變化了解人體生理功能和病理變化,是中醫(yī)學(xué)獨(dú)具特色的診法之一[1]。(剩余16306字)
-
-
- 計(jì)算機(jī)應(yīng)用研究
- 2025年07期
- 多模態(tài)行人重識(shí)別研究綜述...
- 語義通信在邊緣算力網(wǎng)絡(luò)中的應(yīng)用...
- 基于同態(tài)加密和零知識(shí)證明的區(qū)塊...
- HyperledgerFabr...
- PMoE:在P-tuning中...
- 基于大語言模型的多任務(wù)生成式重...
- 基于圖文對(duì)比融合的圖像人物情感...
- 基于深度特征交互與層次化多模態(tài)...
- 反向聚焦細(xì)粒度多模態(tài)語義對(duì)齊的...
- 基于CLIP文本特征增強(qiáng)的剪紙...
- 基于完整超圖神經(jīng)網(wǎng)絡(luò)的捆綁推薦...
- 基于高階鄰域信息交互的自監(jiān)督異...
- 基于超圖和分層頻譜濾波器的序列...
- 針對(duì)圖像指代分割的訓(xùn)練后量化策...
- 基于信息互補(bǔ)與交叉注意力的跨模...
- 基于強(qiáng)化學(xué)習(xí)協(xié)同進(jìn)化算法求解柔...
- 融合實(shí)體鄰域信息的時(shí)序知識(shí)圖譜...
- 互補(bǔ)盲點(diǎn)策略和U型Transf...
- SP-POMDP:堆疊物體抓取...
- 基于果蠅協(xié)同算法求解雙目標(biāo)混裝...
- 優(yōu)化時(shí)間窗改進(jìn)Dijkstra...
- 帶頻繁區(qū)域的空間并置模式挖掘方...
- 輔助任務(wù)增強(qiáng)的知識(shí)追蹤方法...
- 基于沖突避讓的多智能體有效旁路...
- 基于特征融合的音頻偽造檢測(cè)方法...
- 基于多視圖舌象特征融合的中醫(yī)證...
- 多元異構(gòu)耦合網(wǎng)絡(luò)中競爭性輿情信...
- 基于增強(qiáng)控制流圖與孿生網(wǎng)絡(luò)架構(gòu)...
- 獎(jiǎng)勵(lì)回溯DQN驅(qū)動(dòng)的多QoS工...
- 基于QUIC的擁塞控制算法動(dòng)態(tài)...
- CN2Conv:面向物聯(lián)網(wǎng)設(shè)備...
- 面向物流數(shù)據(jù)共享的可撤銷屬性加...
- 一種具有多級(jí)安全目標(biāo)的動(dòng)態(tài)對(duì)稱...
- 基于雙向數(shù)據(jù)流分析與圖抽象嵌入...
- 基于比特切片技術(shù)與指令集的LE...
- 基于隨機(jī)投影與改進(jìn)min-ma...
- 結(jié)合自適應(yīng)局部圖卷積與多尺度時(shí)...
- 基于圖元變換的建筑彩繪紋樣圖像...
- 雙流特征增強(qiáng)與融合的弱監(jiān)督時(shí)序...
- 多尺度降噪自編碼器的遮擋行人重...
- 基于深度正則化的三維高斯人體重...
- 基于雙曲空間的無監(jiān)督視頻異常檢...