李超 楊樞 周瑛
摘要:為了有效降低骨質(zhì)疏松疾病臨床診斷誤診、漏診率,分析骨質(zhì)疏松診斷指標(biāo)體系,綜合粒子群神經(jīng)網(wǎng)絡(luò)和支持向量機(jī)兩種學(xué)習(xí)型算法,以誤差絕對(duì)值和達(dá)到最小為準(zhǔn)則建立疾病診斷分類(lèi)模型。采用蚌埠醫(yī)學(xué)院第一附屬醫(yī)院骨科患者實(shí)際病例數(shù)據(jù)作為樣本集,對(duì)模型進(jìn)行訓(xùn)練和測(cè)試,使誤差達(dá)到規(guī)定要求,并將仿真結(jié)果與單一診斷模型進(jìn)行比較。實(shí)證分析表明組合模型診斷誤差明顯小于單一診斷模型。用基于神經(jīng)網(wǎng)絡(luò)和支持向量機(jī)的線性組合模型診斷原發(fā)性骨質(zhì)疏松病情,是可行有效的方法。
關(guān)鍵詞:神經(jīng)網(wǎng)絡(luò);支持向量機(jī);組合模型;骨質(zhì)疏松診斷
中圖分類(lèi)號(hào):TP183 文獻(xiàn)標(biāo)識(shí)碼:A 文章編號(hào):1009-3044(2014)33-8013-03
Abstract: In order to reduce the misdiagnosis rate of osteoporosis. We analyzed osteoporosis assessment indicator system. Comprehensive application neural networks and support vector machine learning algorithms, established linear combination diagnosis model to achieve the absolute minimum error. Then the combination model was used to compute a large number of actual data in orthopedics Department of first Affiliated Hospital of Bengbu Medical College. And the simulation results were compared with the single diagnostic model. Analysis shows the Diagnostic error of combined model is significantly less than single diagnostic models. The diagnosis of osteoporosis by using combination model based on neural network and support vector machine is feasible and effective.
Key words: neural networks; SVM; combination model; Osteoporosis diagnosis
1 概述
骨質(zhì)疏松癥,在臨床診斷中極易與股骨頭壞死、骨髓瘤、骨軟化癥、骨量不足、遺傳性成骨不全等疾病混淆。世界衛(wèi)生組織以骨量和(或) 骨骼所能承受的力兩個(gè)方面來(lái)評(píng)估骨質(zhì)疏松的程度。但由于骨質(zhì)疏松疾病本身診斷的復(fù)雜性,WHO診斷標(biāo)準(zhǔn)的局限性,及年輕醫(yī)生經(jīng)驗(yàn)不足等原因,導(dǎo)致誤診與漏診情況頻繁出現(xiàn)。作為最具代表性的兩種學(xué)習(xí)型評(píng)價(jià)方法,人工神經(jīng)網(wǎng)絡(luò)和支持向量機(jī)算法近年來(lái)越來(lái)越受到關(guān)注。算法通過(guò)對(duì)樣本的訓(xùn)練和學(xué)習(xí)得到自變量和因變量的內(nèi)在關(guān)系從而能夠模擬專(zhuān)家智慧,評(píng)估或預(yù)測(cè)相關(guān)因素的結(jié)果。神經(jīng)網(wǎng)絡(luò)通過(guò)模擬人腦的結(jié)構(gòu)來(lái)處理信息;而支持向量機(jī)則是一種基于數(shù)據(jù)的統(tǒng)計(jì)學(xué)習(xí)方法。……