Feasibility study of dark-field imaging for colon cancer liver metastases in mice based on a laboratory X-ray grating-based imaging system
摘要
背景与目的:传统X线成像基于X线的吸收特性,对肝脏软组织肿瘤的识别存在一定局限.暗场成像和相衬成像利用X线的散射与折射特性,可提供额外的组织微观结构信息.本研究基于实验室X线光源光栅综合成像系统,初步探索吸收、暗场和相衬3种成像模式在离体小鼠结肠癌肝转移瘤模型中的图像特征及对比度差异,评估该技术的可行性与潜在应用价值.方法:使用3只免疫健全的BALB/c雄性小鼠,经脾注射CT26结肠癌细胞建立肝转移瘤模型,将离体肝脏标本福尔马林固定后,使用实验室X线光源光栅综合成像系统进行扫描,获取数字 X 线摄影(digital radiography,DR)及计算机断层扫描(computed tomography,CT)图像.由4位影像科医师(两位低年资、两位高年资)采用Likert量表独立评估3种CT图像的整体图像质量、肿瘤边缘清晰度以及伪影情况;并计算对比度-噪声比(contrast-to-noise ratio,CNR).鉴于当前属初步探索性实验且有效样本量有限,本研究未进行组间假设检验,结果以描述性统计呈现数值趋势.结果:样本1因成像过程中发生变形未能进行CT重建,仅获得DR图像;样本2和样本3均获得DR及CT图像.吸收DR/CT图像上,肝转移瘤相对于正常肝实质呈等或略高密度,边界不清;暗场DR/CT图像上,肝转移瘤呈低信号,正常肝实质呈高信号,二者对比明显;相衬DR图像上,正常肝实质呈高-低混杂信号,肝转移瘤呈相对均匀信号,相衬CT图像上肝转移瘤显示不清.暗场CT的整体图像质量评分(3.63±0.52)、肿瘤边缘清晰度评分(3.75±0.46)均呈现出高于吸收CT(1.25±0.71;1.63±0.52)和相衬CT(0.38±0.74;0.38±0.52)的数值趋势,伪影评分(1.13±0.35)呈现出低于吸收 CT(2.63±0.74)和相衬CT(3.38±0.74)的数值趋势,观察者间一致性处于中等至良好水平(两位低年资医师间κ=0.709,95%CI:0.512~0.950;两位高年资医师间κ=0.750,95%CI:0.613~0.887;不同年资医师间κ=0.527,95%CI:0.400~0.655).样本2暗场CT的CNR(0.512 8)高于吸收CT(0.290 4)和相衬CT(0.324 1);样本3暗场CT的CNR(0.526 0)亦高于吸收CT(0.198 0)和相衬CT(0.025 9).结论:本研究成功在离体结肠癌肝转移瘤模型中实现暗场CT成像,首次证实其可清晰地区分肝转移瘤与正常肝实质.虽受限于样本规模,但暗场图像在整体图像质量、肿瘤边缘清晰度及伪影控制方面均呈现一致的数值优势,表明其突破传统吸收成像软组织分辨局限的潜力.为后续活体成像的参数优化及效应量估算提供了关键实验依据.
Background and purpose:Conventional X-ray imaging based on absorption has limitations in identifying soft-tissue tumors in the liver.Dark-field and phase-contrast imaging,utilizing X-ray scattering and refraction,can provide additional information on tissue microstructure.This study,based on a laboratory X-ray grating-based integrated imaging system,preliminarily explored the image features and contrast differences among absorption,dark-field,and phase-contrast imaging modes in ex vivo mouse models of colon cancer liver metastases,aiming to assess the feasibility and potential application value of this technique.Methods:Three male BALB/c mice were injected via the spleen with CT26 colon cancer cells to establish liver metastasis models.The excised liver specimens were fixed in formalin and then scanned using the laboratory X-ray grating-based integrated imaging system to obtain digital radiography(DR)and computed tomography(CT)images.Four radiologists(two junior and two senior)independently evaluated the overall image quality,tumor edge sharpness,and artifacts of the three CT modes using a Likert scale,and the contrast-to-noise ratio(CNR)was calculated.Given that this was a preliminary exploratory experiment with a limited effective sample size,no intergroup hypothesis testing was performed;results are presented as descriptive statistics showing numerical trends.Results:Sample 1 could not undergo CT reconstruction because of deformation during imaging,and only DR images were obtained;Samples 2 and 3 both yielded DR and CT images.On absorption DR/CT images,liver metastases showed iso-or slightly high density relative to normal liver parenchyma,with unclear boundaries.On dark-field DR/CT images,liver metastases showed low signal,while normal liver parenchyma showed high signal,with clear contrast between them.On phase-contrast DR images,normal liver parenchyma showed high-low mixed signal,and liver metastases showed relatively uniform signal;On phase-contrast CT images,liver metastases were poorly visualized.The overall image quality score(3.63±0.52)and tumor margin clarity score(3.75±0.46)of dark-field CT showed a numerical trend higher than those of absorption CT(1.25±0.71;1.63±0.52)and phase-contrast CT(0.38±0.74;0.38±0.52),while the artifact score(1.13±0.35)showed a numerical trend lower than those of absorption CT(2.63±0.74)and phase-contrast CT(3.38±0.74).Inter-observer agreement was moderate to good(between the two junior radiologists,κ=0.709,95%CI:0.512-0.950;between the two senior radiologists,κ=0.750,95%CI:0.613-0.887;between radiologists of different seniority,κ=0.527,95%CI:0.400–0.655).The CNR of dark-field CT in Sample 2(0.5128)was higher than that of absorption CT(0.2904)and phase-contrast CT(0.3241);the CNR of dark-field CT in Sample 3(0.5260)was also higher than that of absorption CT(0.1980)and phase-contrast CT(0.0259).Conclusion:This study successfully achieved dark-field CT imaging in ex vivo models of colon cancer liver metastases,demonstrating for the first time that it can clearly distinguish liver metastasis from normal parenchyma.Although limited by sample size,dark-field images consistently showed numerical advantages in overall image quality,tumor edge sharpness,and artifact control,suggesting its potential to overcome the soft-tissue resolution limitations of conventional absorption imaging.This provides key experimental evidence for parameter optimization and effect size estimation in future in vivo imaging studies.