基于图像硬度和边缘纹理特征量的肝纤维化诊断研究-信息处理与通信网络系统专业论文.docx

常肝和非正常肝。(2)利用灰度共生矩阵(GrayLevelCo—occurrenceMatrix，GLCM)计算肝脏CT图像纹理信息，提取肝脏轮廓边缘信息：平均灰度值、灰度方差值和14个边缘纹理特征量，对提取的16个边缘纹理信息进行肝纤维化分类诊断实验。(3)肝纤维化分类诊断实验中采用基于径向基核函数的支持向量机(SupportVectorMachine，SVM)算法，采用时间和成本代价最高的穷举遍历法和留一法循环遍历所有特征组合进行分类诊断实验。实验结果表明当边缘纹理特征量数目为3至7个时，分类实验有最佳诊断效果；分析边缘纹理特征量的权重排序发现：最大相关系数、对比度、共生差方差、差异平方和和共生和方差在分类实验中的权重值远远大于其它特征量。(4)比较边缘纹理特征量和纹理特征量、形状特征量在肝纤维化分类诊断中的实验效果，比较特征量提取方法的优劣性，分析特征量权重分布特点。研究基于图像硬度和边缘纹理特征量的肝纤维化混合诊断实验，混合实验大大提高了诊断的整体平均准确率。关键词：计算机辅助诊断薄板样条灰度共生矩阵边缘纹理特征量支持向量机万方数据DIAGNOSISOFHEI’ATICFIBRoSISBASEDoNIMAGEHARDNESSAN’DEDGETEXTUREFEAT．URESABSTRACTLivercirrhosisisacommonclinicalchronicliverdisease，Itiscausedbyalong—termvarietyoffactors，andisadiffuselivedamage．Liverfibrosisistheearlyclinicalpathologysymptomoflivercirrhosis，andhasbeenfoundisreversible．Therefore，earlydetectionandtimelytreatmentoflivefibrosiscansignificantlyreducethemortalityrateofliverdisease．Atpresent，thediagnosisoflivercirrhosisandliverfibrosismainlyrelyonbiopsymethods．Thisinvasiveexaminationnotonlydamagesthephysicalandmentalhealthofpatients，butalsotheclinicalexperienceofdoctorswilldirectlyaffecttheresultsofdiagnosis．Thusitisurgenttofindamethodofearlynon—invasivediagnosisofliverfibrosis．、胍ththedevelopmentofmedicaltechnologyandthedigitalimageprocessingtechnology,thecomputer—aideddiagnosis(CAD)technologyisintroducedbypeopleinthediagnosisofliverfibrosis，greatlyimprovingthediagnosticefficiencyandaccuracy．Accordingtothecharacteristicsoftheshapefeaturesistime—consumingandpoorscalabilityanddiagnosticeffectoftexturefeaturesisnotgoodinthediagnosisofhepaticfibrosis，SOinthispaper,a万方数据researchofcomputer-aideddiagnosisofhepaticfibrosisonimagehardnessandedgetexturefeaturesisproposed，theprimarycoverageofthestudyisarranged：(1)Analyzingthesoftnessoftheliverbyimageprocessingbasedonthegridshapedliverdeformationimage．InspatialdomainThinPlatespline(TPS)algorithmisusedtocalculatetheminimumbendingenergyofdeformation；inthefrequencydomain，FastFourierTransform(FFT)algorithmisusedtocalculatethespectralvaluesofthecharacteris