人工神经网络肝癌CT影像辅助诊断模型的建立-实用放射学杂志.PDFVIP

2006 9 22 9 J PractRadio,l Sep 2006, Vol. 22, No. 9 1079 CT 1 1 1 1 2 3 4 张 波 ,张治英 ,徐德忠, 闫永平 ,彭 勇 , 夏结来 ,秦明新 5 2 1 6 2 宋振顺 ,高 凯,姜建辉 ,贾 娟 , 李 琼 ( 1, 710032; 2; 3; 4; 5; 6 ) : CT ( 110)( 123) CT, CT CT (back propagation, BP), , , CT , 12, , , CT 98 96 , 97 CT : ; ;, X : R735. 7; R814. 42 : A : 1002- 1671( 2006) 09- 1079- 04 Estab lishm ent of CT Im aging A ccessory D iagnostic M odel Based on A rtificial N euralN etwork in H epatocellu lar Carc inom a ZHAN G B o, ZHANG Zhi-y ing , X U D e - zhong, YAN Yong -p ing, PEN G Yong, X ia J ie - lai, Q IN M ing -x in, SONG Zhen- hun, GA O K ai, J IANG J ian- hui, J IA Juan, LI Q iong (D ep artm en t of Ep idem iology of P rev entiveM edicine, the Fourth M ilitary M ed ical Univ er ity, X i an 710032, Ch ina) Abstrac t Objec tive To establishCT accessory diagnosticmodelbased on artificialneuralnewt ork (ANN) in hepatocellularcarcinoma (HCC).M e thods CT mi ages ofHCC ( 110 cases) and other liveroccupied disease (123 cases) were preprocessed. CT features ofeachCT mi agewere extracted that actecd as the training and evaluating database. CT accessory diagnosticmodel ofANN for diagnosis ofHCCwas constructedw ith aback- propagation(BP). After trained the newt ork with traning database, the evaluating databasewas through thenetwork. Thesensitivity, specificity and accuracyof thenewt ork in diagnosingHCC according to theoutput data. R esults Thenetwork trainingprocess showed that the training lasted 12 epoch. The classification ofdatabasewas proper. Effectof training on newt orkwas verywel.l The sensitivity, specificity and accuracy ofANN were98 , 96 and97 respectively fordiagnosis ofCT diagnosis ofHCC. Con clusion TheAN