資源描述:
《基于神經(jīng)網(wǎng)絡(luò)的肝臟b超圖像識(shí)別研究》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、西南科技大學(xué)碩士研究生學(xué)位論文第1頁(yè)摘要根據(jù)肝臟B超圖像進(jìn)行脂肪肝的診斷,是病變確診的主要方法。由于B超圖像的質(zhì)量較差,目前的診斷以定性為主,受主觀因素影響較大。因此,研究肝臟B超圖像的紋理特征,獲取量化參數(shù),采用計(jì)算機(jī)輔助手段進(jìn)行分類識(shí)別,有助于提高臨床診斷的準(zhǔn)確性及效率。本文著重對(duì)肝臟B超圖像的特征提取和分類識(shí)別兩大中心問(wèn)題進(jìn)行了研究和仿真。特征提取部分以圖像預(yù)處理為基礎(chǔ),采用灰度共生矩陣法,對(duì)正常肝和脂肪肝B超圖像進(jìn)行了特征值的計(jì)算。在對(duì)特征參數(shù)進(jìn)行了比較和選擇后,確定了最能反映圖像特征的參數(shù)組合:能量(角二階矩)、熵和反差分矩。在分類識(shí)別部分,利
2、用Matlab的神經(jīng)網(wǎng)絡(luò)工具箱設(shè)計(jì)了自組織特征映射(sOFM)和誤差反向傳播(BP)兩種神經(jīng)網(wǎng)絡(luò)分類器。采用了兩種樣本選擇方案對(duì)網(wǎng)絡(luò)進(jìn)行訓(xùn)練和仿真,取得了較好的識(shí)別效果。訓(xùn)練好的分類器既能識(shí)別出正常肝和脂肪肝,又能對(duì)脂肪肝的嚴(yán)重程度做出分類。本文建立的識(shí)別方法有助于醫(yī)生快速、準(zhǔn)確地對(duì)肝臟B超圖像作出判斷識(shí)別。關(guān)鍵詞:脂肪肝B超圖像灰度共生矩陣特征提取神經(jīng)網(wǎng)絡(luò)西南科技大學(xué)碩士研究生學(xué)位論文第1I頁(yè)AbstractThediagnosisaboutfattyliveronthebaseofultrasoundimageisanimponantmethod.H
3、oweVer,thequalityofultrasoundimagesisrelativelyp00r.Atpresent,thediagnosisabOutfattyliverisdeterminedbythenature,andtheconclusionofdiseaseisofteninnuencedbydoctors.Therefbre,itishelpfultostudythetexturalcharactersofliverB-Scanultrasonicimage,thequaIltmcationfeaturecharacters,aIld
4、recognizingtheseimagesbycomputerassistancewaystoincreasetheaccllracyande衢ciencyofclinicaldiagnosis.Inthispaper,bothfeatureextractionandclassincationrecognitionoffatty1iverB·Scanultrasonicimage8remainlystudiedaIldsimulated.Inthepartoffeatureextraction,mepapercalculatesthefeatureso
5、fnormalliverandfattyliVer’sB-ScanultrasonicimagesbyusingtllemethodofGray—LevelCo—occurreneeMatricesonthebaseofimagepreprocess.Anerfeaturescomparison柚dselection,wedeterminethebestfcaturescombination,includingangularsecondmoment,entropy柚dinversedifferentialmoment.Inthepanofclassinc
6、ationrecognition,wedesignSOFMandBPneuralnetworkclassifiersbyusingtheneuralnetwork100lboxofMatlab.Usingtwokindsofs鋤pleselectionplanstoexerciseandsimulatetheclassifiersandobtaininggoodrecognitioneff色ct.TheexercisedneuralnetworkclassifiersnotonlycanrecognizetheB-Scanimagesofnormalli
7、verandfattyliverbutalsocanfunherclassifyitsdegree0fseriousness.TherecognitionmethodofthispapercanhelpdoctorstodiscernB—Scanimageofliverquicklyalldaccurately.KeyWords:f8nyliVer;B—Scaninlage;gray—levelco—oeeurrencematrix;featureextraction:neufalnetwork獨(dú)創(chuàng)性聲明本人聲明所呈交的論文是我個(gè)人在導(dǎo)師指導(dǎo)下進(jìn)行的研究
8、工作及取得的研究成果。盡我所知,除了文中特別加以標(biāo)注和致謝的地方外,論文中不包含