论著摘要 |【CT】滤波和噪声对头颈癌患者CT和CBCT影像组学特征鲁棒性的影响(双语版)

2017-09-28 10:37:14 admin 3

On the Impact of Smoothing and Noise on Robustness of CT and CBCT Radiomics Features for Patients with Head and Neck Cancers.

发表日期: 2017.04.17   来源: Medical Physics, 2017,44,(5): 1755–1770.

作者

Hassan Bagher-Ebadian1,2, Farzan Siddiqui1, Chang Liu1, Benjamin Movsas1, Indrin J. Chetty1

作者介绍:

1. Department of Radiation Oncology, Henry Ford Hospital, Detroit, MI, USA

2. Department of Physics, Oakland University, Rochester, MI, USA

 

摘要

Abstract

目的

我们调查了采用从18例分级放射治疗的口咽癌患者中获得的断层CTpCT)和锥束CTCBCT)图像数据提取的影像组学特征的特性。图像经过滤波,锐化和噪声,以评估相对于基线数据集的特征的变化。

Purpose

We investigated the characteristics of radiomics features extracted from planning CT (pCT) and cone beam CT (CBCT) image datasets acquired for 18 oropharyngeal cancer patients treated with fractionated radiation therapy. Images were subjected to smoothing, sharpening, and noise to evaluate changes in features relative to baseline datasets.

 

方法

从肿瘤体积的pCTCBCT图像中中提取纹理特征,根据以下八个不同类别进行轮廓分析:基于强度的直方图特征(IBHF),灰度运行长度(GLRL),劳氏纹理信息(LAWS),离散正交S变换(DOST),局部二元模式(LBP),二维小波变换(2DWT),二维加博滤波(2DGF)和灰度共生矩阵(GLCM)。总共提取了165个影像组学特征。使用具有不同信噪比(SNR = 5,10,15,20,25,35,50,75,100150)的高斯噪声模型在特征提取之前对图像进行后处理。图像数据集中使用具有不同截止频率(从0.04580.7321个周期-mm-1变化的)的高斯滤波器。噪声和滤波对每个提取特征的影响使用后处理和基线图像上各个值之间的平均绝对百分比变化(MAPC)进行量化。Fisher方法联合韦尔奇P值用于显著性检验。对比3处研究:(a)基线pCT对比改进的pCT(使用给定的滤波器);(b)基线CBCT对比改进的CBCT;(c)基线和改进的pCT对比基线和改进的CBCT

Methods

Textural features were extracted from tumor volumes, contoured on pCT and CBCT images, according to the following eight different classes: intensity based histogram features (IBHF), gray level run length (GLRL), law's textural information (LAWS), discrete orthonormal stockwell transform (DOST), local binary pattern (LBP), two-dimensional wavelet transform (2DWT), Two dimensional Gabor filter (2DGF), and gray level co-occurrence matrix (GLCM). A total of 165 radiomics features were extracted. Images were post-processed prior to feature extraction using a Gaussian noise model with different signal-to-noise-ratios (SNR = 5, 10, 15, 20, 25, 35, 50, 75, 100, and 150). Gaussian filters with different cut off frequencies (varied discreetly from 0.0458 to 0.7321 cycles-mm−1) were applied to image datasets. Effect of noise and smoothing on each extracted feature was quantified using mean absolute percent change (MAPC) between the respective values on post-processed and baseline images. The Fisher method for combining Welch P-values was used for tests of significance. Three comparisons were investigated: (a) Baseline pCT versus modified pCT (with given filter applied); (b) Baseline CBCT versus modified CBCT, and (c) Baseline and modified pCT versus baseline and modified CBCT.

结果

CTCBCT图像数据集提取的特征对低通滤波(MAPC = 17.5%, math formula= 0.93CBCTMAPC = 7.5%,  math formula= 0.98pCT)和噪声(MAPC = 27.1%, math formula= 0.89CBCTMAPC = 34.6%, math formula= 0.61pCT)具有鲁棒性。提取的特征对高通滤波呈现显著影响(MAPC = 187.7%,CBCT  math formula<0.0001,对于pCTMAPC = 180.6%, math formula<0.01)。对于pCT,最受低通滤波影响的特征为:对于pCT数据集,LAWSMAPC = 11.2%, math formula= 0.44),GLRLMAPC = 9.7%, math formula= 0.70)和IBHFMAPC = 21.7%, math formula= 0.83);对于CBCT图像数据集,LAWSMAPC = 20.2%, math formula= 0.24),GLRLMAPC = 14.5%, math formula= 0.44),2DCTMAPC = 16.3%, math formula= 0.52)。对于pCT数据集,最受噪声影响的特征是GLRLMAPC = 29.7%,  math formula= 0.06),LAWSMAPC = 96.6%, math formula= 0.42),GLCMMAPC = 36.2%,  math formula= 0.48);而LBPFMAPC = 5.2%, math formula= 0.99)对噪声相对不敏感。对于CBCT数据集,GLRLMAPC = 8.9%, math formula= 0.80)和LAWSMAPC = 89.3%, math formula= 0.81)特征受到噪声的影响,而LBPFMAPC = 2.2%,  math formula= 0.99)和DOSTMAPC = 13.7%,  math formula= 0.98)的特征对噪声不敏感。除了这15个特征,从基线pCTCBCT图像数据集中提取的剩余150个纹理特征都没有观察到显著差异(MAPC = 90.1%,  math formula= 0.26)。

Results

Features extracted from CT and CBCT image datasets were robust to low-pass filtering (MAPC = 17.5%,  math formula = 0.93 for CBCT and MAPC = 7.5%,   math formula = 0.98 for pCT) and noise (MAPC = 27.1%,   math formula=  0.89 for CBCT, and MAPC = 34.6%,  math formula = 0.61 for pCT). Extracted features were significantly impacted (MAPC=187.7%,   math formula < 0.0001 for CBCT, and MAPC = 180.6%,   math formula < 0.01 for pCT) by LOG which is classified as a high-pass filter. Features most impacted by low pass filtering were LAWS (MAPC = 11.2%,   math formula = 0.44), GLRL (MAPC = 9.7%,   math formula = 0.70) and IBHF (MAPC = 21.7%,   math formula = 0.83), for the pCT datasets, and LAWS (MAPC = 20.2%,   math formula = 0.24), GLRL (MAPC = 14.5%,   math formula = 0.44), and 2DGF (MAPC=16.3%,   math formula = 0.52), for CBCT image datasets. For pCT datasets, features most impacted by noise were GLRL (MAPC = 29.7%,   math formula = 0.06), LAWS (MAPC = 96.6%,   math formula = 0.42), and GLCM (MAPC = 36.2%,   math formula = 0.48), while the LBPF (MAPC = 5.2%,   math formula = 0.99) was found to be relatively insensitive to noise. For CBCT datasets, GLRL (MAPC = 8.9%,   math formula = 0.80) and LAWS (MAPC = 89.3%,   math formula = 0.81) features were impacted by noise, while the LBPF (MAPC = 2.2%,   math formula = 0.99) and DOST (MAPC = 13.7%,   math formula = 0.98) features were noise insensitive. Apart from 15 features, no significant differences were observed for the remaining 150 textural features extracted from baseline pCT and CBCT image datasets (MAPC = 90.1%,   math formula = 0.26).

结论

针对头颈部癌症患者的断层CT和日常CBCT图像数据提取的影像组学特征对于低功率高斯噪声和低通滤波是有鲁棒性的,但受到高通滤波的影响。从CBCTpCT图像数据集提取的纹理特征是相似的,表明pCTCBCT用于研究影像组学特征作为预后的可能生物标志物具有互换性。

Conclusions

Radiomics features extracted from planning CT and daily CBCT image datasets for head/neck cancer patients were robust to low-power Gaussian noise and low-pass filtering, but were impacted by high-pass filtering. Textural features extracted from CBCT and pCT image datasets were similar, suggesting interchangeability of pCT and CBCT for investigating radiomics features as possible biomarkers for outcome.

 

阅读原文:10.1002/mp.12188


慧影医疗科技(北京)有限公司

地点:北京市海淀区中关村东升科技园B2-C103

电话:400-890-9020

邮箱:radcloud@huiyihuiying.com

关闭
图片
图片
  • 深度学习科研云平台