本文選題：頻域光學(xué)相干層技術(shù)　切入點(diǎn)：青光眼　出處：《電子與信息學(xué)報(bào)》2017年05期

【摘要】：頻譜域光學(xué)相干層析技術(shù)是一種廣泛應(yīng)用于眼科疾病診斷的成像技術(shù),而視網(wǎng)膜層分割對(duì)青光眼的診斷有很好的參考價(jià)值。該文利用隨機(jī)森林分類器尋找視網(wǎng)膜層間單像素寬的邊界,隨機(jī)森林分類器由12個(gè)特征訓(xùn)練產(chǎn)生,其中相對(duì)灰度特征和鄰域特征較好地解決灰度不均勻的分割誤差大問(wèn)題。對(duì)10組帶有青光眼病變的視網(wǎng)膜圖像進(jìn)行分割,并與傳統(tǒng)算法和Iowa軟件進(jìn)行比較,平均邊界絕對(duì)誤差為9.20±2.57μm,11.33±2.99μm和10.27±3.01μm。實(shí)驗(yàn)結(jié)果表明,改進(jìn)算法可以較好地分割視網(wǎng)膜神經(jīng)纖維層。
[Abstract]:Spectral domain optical coherence tomography (OCS) is a widely used imaging technique for the diagnosis of ophthalmic diseases. The retinal layer segmentation has a good reference value for the diagnosis of glaucoma. In this paper, a random forest classifier is used to find the boundary between the retinal layers with a single pixel width. The random forest classifier is generated by 12 feature training. The relative grayscale feature and neighborhood feature can solve the problem of big error of uneven grayscale segmentation. Ten groups of retinal images with glaucoma are segmented and compared with the traditional algorithm and Iowa software. The average boundary absolute error is 9.20 鹵2.57 渭 m 11.33 鹵2.99 渭 m and 10.27 鹵3.01 渭 m respectively. The experimental results show that the improved algorithm can segment the retinal nerve fiber layer well.
【作者單位】： 南京理工大學(xué)計(jì)算機(jī)科學(xué)與工程學(xué)院;福建省信息處理與智能控制重點(diǎn)實(shí)驗(yàn)室(閩江學(xué)院);濟(jì)南大學(xué)信息科學(xué)與工程學(xué)院;
【基金】：國(guó)家自然科學(xué)基金(61671242) 中央高�；究蒲袠I(yè)務(wù)費(fèi)專項(xiàng)資金(30920140111004) 六大人才高峰(2014-SWYY-024) 福建省信息處理與智能控制重點(diǎn)實(shí)驗(yàn)室(閩江學(xué)院)開(kāi)放課題基(MJUKF201706)~~
【分類號(hào)】：TP391.41;R770.4
，

本文編號(hào)：1674787