-
乳腺专用γ显像(breast specific gamma imaging,BSGI)是一项高诊断效能的新技术,主要用于致密性乳腺及超声、钼靶诊断不明确的乳腺病例[1]。作为补充显像,其灵敏度优于钼靶,特异度优于超声和MRI[2-3]。中国女性患有致密性乳腺的比例较高,BSGI的诊断价值尤为突出。
根据2010年美国核医学会发布的《BSGI操作指南》1.0版[4],乳腺影像报告和数据系统(breast imaging reporting and data system,BIRADS)主要参考病灶的放射性摄取程度、病灶形态及边缘进行诊断。已有部分研究者提出,基于上述标准,BSGI面临假阳性较高且特异性受限的问题[5-7]。Tan等[8]和Park等[9]曾分别使用最大肿瘤/非肿瘤比值(the ratio of tumor to non-tumor, T/NT)以及延迟相显像剂的洗脱,提高病灶诊断的特异度,但暂未进行病灶灰度分布特征的相关研究。本研究回顾性分析乳腺病灶的BSGI图像特征,拟探讨有利于鉴别诊断乳腺病变的潜在图像特征。
-
行BSGI且乳腺病灶经手术病理确诊的女性患者有282例,其中10例为已确诊乳腺恶性肿瘤,且BSGI检查前已行手术或放化疗,故予以剔除。本研究最终入组272例患者,其中包括临床症状可疑者25例、超声和(或)钼靶诊断模棱两可及可疑恶性者224例、病理确诊但未行治疗者8例、乳腺癌高危人群7例、因致密性乳腺行筛查者8例。患者年龄为20~89(51.63±12.43)岁,共293个病灶,包括187个恶性病灶(187/293,63.8%)和106个良性病灶(106/293,36.2%)(表1)。187个恶性病灶的最大长径为0.5~11.0 cm,其中0.5~1.0 cm 17个、1.1~1.5 cm 43个、1.6~2.0 cm 36个、>2.0 cm 91个。
病灶类型 病灶数(个) 病理分型 数量(个) 恶性病灶 187 浸润性导管癌 135 导管原位癌 32 浸润性小叶癌 3 小叶原位癌 4 浸润性乳头状癌 3 乳头状原位癌 3 恶性分叶状肿瘤 1 黏液癌 2 神经内分泌肿瘤 4 良性病灶 106 纤维腺瘤 49 乳腺腺病 36 导管内乳头状瘤 9 慢性或急性感染 6 良性分叶状肿瘤 1 乳腺导管上皮不典型增生 4 间质胶原化 1 表 1 入选病例的293个乳腺病灶的病理分型
Table 1. Pathological subtypes of the included 293 breast lesions
-
病灶最大T/NT与病理结果的ROC曲线提示:ROC曲线下面积为0.802(标准误:0.026;95%CI:0.750~0.853,P=0.000)。本研究中病灶最大T/NT的最佳界值为1.75 (图1)。
图 1 乳腺病灶最大肿瘤/非肿瘤比值与病理结果的受试者工作特征曲线
Figure 1. Receiver operating characteristic curve between the maximum ratio of tumor to non-tumor and pathology for determining the best cut-off ratio of tumor to non-tumor in the diagnosis of breast cancer
二变量秩相关分析结果显示,病灶形态、病灶边缘是否清晰、灰度分布是否存在“偏心核心”(图2)、BIRADS结果及最大T/NT(界值:1.75)都与病理结果存在显著的相关性(Spearman相关系数分别为0.551、0.371、0.290、0.489和0.436,均P=0.000),即结节状病灶、病灶边缘清晰、灰度分布缺乏“偏心核心”、BIRADS 4~5级以及最大T/NT >1.75均与恶性病灶呈正相关。
图 2 乳腺病灶“偏心核心”的典型BSGI图 患者女性,33岁,发现左乳肿块1年,母亲有乳腺癌病史,CA125为52.10 U/mL,AFP、CA199、CEA均为(−),超声提示良性病变可能,大小约为1.7 cm×0.7 cm。图中,左乳结节状放射性异常浓聚灶,并见“偏心核心”(A和C,红色箭头),左侧头尾位(A和C)和左侧内外侧斜位(B和D)的最大肿瘤/非肿瘤比值分别为2.0和1.6。BSGI诊断为BIRADS 5级,考虑为恶性病变可能性大;术后病理为乳腺纤维腺瘤。BSGI:乳腺专用γ显像;CA:糖类抗原;AFP:甲胎蛋白;CEA:癌胚抗原;BIRADS:乳腺影像报告和数据系统。
Figure 2. Characteristic image of "a decentered core" of the breast lesion in breast specific gamma imaging
二分类Logistic回归分析结果显示,病灶形态(OR=0.013,95%CI:3.664~21.846,P=0.000)、病灶边缘是否清晰(OR=2.121,95%CI:1.061~4.239,P=0.033)以及灰度分布是否存在“偏心核心”(OR=12.927,95%CI:5.415~30.863,P=0.000)均与病理结果的相关性有统计学意义,而BIRADS结果(OR=2.546,95%CI:0.944~6.867,P=0.065)和最大T/NT(界值:1.75)(OR=1.570,95%CI:0.659~3.738,P=0.308)均与病理结果的相关性无统计学意义。
-
BIRADS结果、最大T/NT、病灶形态、病灶边缘是否清晰、灰度分布是否存在“偏心核心”对乳腺病变的诊断效能见表2。其中,病灶形态、病灶边缘是否清晰、灰度分布是否存在“偏心核心”对乳腺病变的合并诊断灵敏度、特异度和准确率分别为88.2%(165/187)、81.1%(86/106)和85.7%(251/293)。
诊断标准 灵敏度 特异度 准确率 阳性预测值 阴性预测值 BIRADS结果 88.2%(165/187) 58.5%(62/106) 77.5%(227/293) 78.9%(165/209) 73.8%(62/84) 最大T/NT(界值:1.75) 76.5%(143/187) 67.9%(72/106) 73.4%(215/293) 80.8%(143/177) 62.1%(72/116) 病灶形态 92.0%(172/187) 58.5%(62/106) 79.9%(234/293) 79.6%(172/216) 80.5%(62/77) 病灶边缘是否清晰 66.8%(125/187) 71.7%(76/106) 68.7%(201/293) 80.6%(125/155) 55.1%(76/138) 灰度分布是否存在“偏心核心” 95.7%(179/187) 27.4%(29/106) 71.0%(208/293) 69.9%(179/256) 78.4%(29/37) 病灶形态+病灶边缘是否清晰
+灰度分布是否存在“偏心核心”88.2%(165/187) 81.1%(86/106) 85.7%(251/293) 89.2%(165/185) 79.6%(86/108) 注:表中,BSGI:乳腺专用γ显像;BIRADS:乳腺影像报告和数据系统;T/NT:肿瘤/非肿瘤比值。 表 2 BSGI的图像特征中不同诊断标准对乳腺病变的诊断效能
Table 2. Diagnostic efficacy of the different diagnostic standards in image feaures of breast specific gamma imaging
不同诊断标准的ROC曲线参见图3。图中,BIRADS结果的曲线下面积为0.731(标准误:0.033;95%CI:0.667~0.795;P=0.000);最大T/NT(界值:1.75)的曲线下面积为0.722(标准误:0.032;95%CI:0.659~0.785;P=0.000);病灶形态、病灶边缘是否清晰、灰度分布是否存在“偏心核心”的曲线下面积分别为0.752(标准误:0.032;95%CI:0.689~0.815;P=0.000)、0.693 (标准误:0.032;95%CI:0.630~0.756;P=0.000)和0.603 (标准误:0.036;95%CI:0.533~0.673;P=0.003)。病灶形态+病灶边缘是否清晰+灰度分布是否存在“偏心核心”合并诊断的曲线下面积为0.847(标准误:0.026;95%CI:0.796~0.898;P=0.000)。上述ROC曲线的Z检验结果提示:BIRADS结果与最大T/NT(界值:1.75)的诊断效能差异无统计学意义(Z=0.379,P=0.704);病灶形态与病灶边缘是否清晰的诊断效能差异无统计学意义(Z=1.942,P=0.052);灰度分布是否存在“偏心核心”与病灶形态、病灶边缘是否清晰的诊断效能差异均有统计学意义(Z=3.574、2.316,均P<0.05);而病灶形态、病灶边缘是否清晰、灰度分布是否存在“偏心核心”三者联合诊断的效能均优于三者单独诊断的效能(Z=4.380、4.816和9.275,均P<0.05),且优于BIRADS结果和最大T/NT(界值:1.75)的单独诊断效能(Z=4.079、4.090,均P<0.05)。
-
基于BIRADS的独立诊断,本研究293个病灶中有66个病灶被误诊,包括22个假阴性病灶[7个低级别浸润性导管癌、10个导管原位癌、1个黏液癌、1个恶性分叶状肿瘤和3个浸润性小叶癌(上述病灶最大径均>1 cm)]和44个假阳性病灶(22个纤维腺瘤、8个乳腺腺病、5个导管内乳头状瘤、5个感染性病变、2个乳腺导管上皮不典型增生、1个良性分叶状肿瘤和1个间质胶原化)。
基于病灶形态+病灶边缘是否清晰+灰度分布是否存在“偏心核心”3个图像特征的联合诊断,本研究293个病灶中有42个病灶被误诊,包括22个假阴性病灶[9个浸润性导管癌、7个导管原位癌、1个浸润性小叶癌、1个浸润性乳头状癌、1个小叶原位癌、1个恶性分叶状肿瘤、1个黏液癌和1个神经内分泌肿瘤(上述病变最大径均>1 cm)]和20个假阳性病灶(12个纤维腺瘤、4个乳腺腺病、1个导管内乳头状瘤、1个间质胶原化和2个乳腺导管上皮不典型增生)。
乳腺专用γ显像的图像特征对乳腺病变的鉴别诊断价值
Efficacy of the image features of breast-specific Gamma imaging in the differential diagnosis of breast lesions
-
摘要:
目的 探讨乳腺专用γ显像(BSGI)的图像特征对乳腺病变的鉴别诊断价值。 方法 回顾性分析2014年7月至2015年6月行BSGI的272例中国女性乳腺疾病患者(共293个病灶),观察BSGI图像上病灶形态、病灶边缘是否清晰、灰度分布是否存在“偏心核心”、乳腺影像报告和数据系统(BIRADS)结果以及最大肿瘤/非肿瘤比值(T/NT)。采用二变量秩相关分析及二分类Logistic回归分析法计算上述特征与病理结果的相关性。基于病灶计算所有图像特征的独立诊断效能以及上述显著相关特征的合并诊断效能,用MedCalc软件行Z检验比较上述特征的受试者工作特征曲线。 结果 病灶形态(OR=0.013,95%CI:3.664~21.846,P=0.000)、病灶边缘是否清晰(OR=2.121,95%CI:1.061~4.239,P=0.033)以及灰度分布是否存在“偏心核心”(OR=12.927,95%CI:5.415~30.863,P=0.000)与病理结果显著相关。三者的灵敏度、特异度分别为92.0%(172/187)和58.5%(62/106)、66.8%(125/187)和71.7%(76/106)、95.7%(179/187)和27.4%(29/106)。三者合并诊断效能最佳,灵敏度、特异度、阳性预测值、阴性预测值、准确率分别为88.2%(165/187)、81.1%(86/106)、89.2%(165/185)、79.6%(86/108)和85.7%(251/293),较BIRADS以及最大T/NT(界值:1.75)更准确,且差异均有统计学意义(Z=4.079、4.090,均P<0.05)。 结论 病灶形态、病灶边缘是否清晰以及灰度分布是否存在“偏心核心”可作为BSGI鉴别诊断乳腺病灶的图像特征,三者联合诊断能提高BSGI在乳腺病变中的独立诊断价值。 Abstract:Objective To investigate the image features of breast specific gamma imaging(BSGI) in the differential diagnosis of breast lesions. Methods A total of 272 Chinese female patients(including 293 lesions) who underwent BSGI between July 2014 to June 2015 were included. Several characteristics of the shape of the lesion, clarity of the boundary, grey scale distribution(the existence of a decentered core), breast imaging reporting and data system(BIRADS), and maximum tumor to non-tumor ratio(T/NT) were recorded. The correlation of each feature with the pathology was evaluated by rank correlation analysis and binary logistic regression analysis. All features were used in the diagnosis of the 293 lesions. Each independent lesion-based diagnostic performance as well as the combined diagnostic efficacy of statistically significant features were evaluated. Using MedCalc software, Z test based on receiver operating characteristic curve was used between each pair of image features to figure out their possible differences. Results Three imaging features including the shape of the lesion(OR=0.013, 95%CI: 3.664–21.846, P=0.000), the clarity of boundary(OR=2.121, 95%CI: 1.061–4.239, P=0.033), and grey scale distribution(the existence of a decentered core)(OR=12.927, 95%CI: 5.415–30.863, P=0.000) were significantly related with the pathology. The sensitivity and specificity of the three former characteristics were 92.0%(172/187) and 58.5%(62/106), 66.8%(125/187) and 71.7%(76/106), and 95.7%(179/187) and 27.4%(29/106), respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the combined diagnosis of the three former image characteristics were 88.2%(165/187), 81.1%(86/106), 89.2%(165/185), 79.6%(86/108), and 85.7%(251/293), respectively. With the best performance of all, this combined diagnosis has a higher diagnostic performance than BIRADS and maximum T/NT(cutoff ratio:1.75)(Z=4.079 and 4.090, both P<0.05). Conclusions The shape of the lesion, the clarity of boundary, and the grey-scale distribution(the existence of a decentered core) could be three important differential diagnostic standards of breast lesions in BSGI. With the combined diagnosis of the three features, the efficacy of independent diagnosis of BSGI in breast lesions could be improved. -
Key words:
- Breast diseases /
- Radionuclide imaging /
- Mammography /
- Diagnosis, differential /
- Image features
-
图 2 乳腺病灶“偏心核心”的典型BSGI图 患者女性,33岁,发现左乳肿块1年,母亲有乳腺癌病史,CA125为52.10 U/mL,AFP、CA199、CEA均为(−),超声提示良性病变可能,大小约为1.7 cm×0.7 cm。图中,左乳结节状放射性异常浓聚灶,并见“偏心核心”(A和C,红色箭头),左侧头尾位(A和C)和左侧内外侧斜位(B和D)的最大肿瘤/非肿瘤比值分别为2.0和1.6。BSGI诊断为BIRADS 5级,考虑为恶性病变可能性大;术后病理为乳腺纤维腺瘤。BSGI:乳腺专用γ显像;CA:糖类抗原;AFP:甲胎蛋白;CEA:癌胚抗原;BIRADS:乳腺影像报告和数据系统。
Figure 2. Characteristic image of "a decentered core" of the breast lesion in breast specific gamma imaging
表 1 入选病例的293个乳腺病灶的病理分型
Table 1. Pathological subtypes of the included 293 breast lesions
病灶类型 病灶数(个) 病理分型 数量(个) 恶性病灶 187 浸润性导管癌 135 导管原位癌 32 浸润性小叶癌 3 小叶原位癌 4 浸润性乳头状癌 3 乳头状原位癌 3 恶性分叶状肿瘤 1 黏液癌 2 神经内分泌肿瘤 4 良性病灶 106 纤维腺瘤 49 乳腺腺病 36 导管内乳头状瘤 9 慢性或急性感染 6 良性分叶状肿瘤 1 乳腺导管上皮不典型增生 4 间质胶原化 1 表 2 BSGI的图像特征中不同诊断标准对乳腺病变的诊断效能
Table 2. Diagnostic efficacy of the different diagnostic standards in image feaures of breast specific gamma imaging
诊断标准 灵敏度 特异度 准确率 阳性预测值 阴性预测值 BIRADS结果 88.2%(165/187) 58.5%(62/106) 77.5%(227/293) 78.9%(165/209) 73.8%(62/84) 最大T/NT(界值:1.75) 76.5%(143/187) 67.9%(72/106) 73.4%(215/293) 80.8%(143/177) 62.1%(72/116) 病灶形态 92.0%(172/187) 58.5%(62/106) 79.9%(234/293) 79.6%(172/216) 80.5%(62/77) 病灶边缘是否清晰 66.8%(125/187) 71.7%(76/106) 68.7%(201/293) 80.6%(125/155) 55.1%(76/138) 灰度分布是否存在“偏心核心” 95.7%(179/187) 27.4%(29/106) 71.0%(208/293) 69.9%(179/256) 78.4%(29/37) 病灶形态+病灶边缘是否清晰
+灰度分布是否存在“偏心核心”88.2%(165/187) 81.1%(86/106) 85.7%(251/293) 89.2%(165/185) 79.6%(86/108) 注:表中,BSGI:乳腺专用γ显像;BIRADS:乳腺影像报告和数据系统;T/NT:肿瘤/非肿瘤比值。 -
[1] Schillaci O, Buscombe JR. Breast scintigraphy today: indications and limitations[J]. Eur J Nucl Med Mol Imaging, 2004, 31 Suppl 1: S35−45. DOI: 10.1007/s00259−004−1525−x [2] Yu X, Hu G, Zhang Z, et al. Retrospective and comparative analysis of 99mTc-Sestamibi breast specific gamma imaging versus mammography, ultrasound, and magnetic resonance imaging for the detection of breast cancer in Chinese women[J/OL]. BMC Cancer, 2016, 16: 450[2018-05-05]. https://bmccancer.biomedcentral.com/articles/10.1186/s12885-016-2537-1. DOI: 10.1186/s12885-016-2537-1. [3] Cho MJ, Yang JH, Yu YB, et al. Validity of breast-specific gamma imaging for Breast Imaging Reporting and Data System 4 lesions on mammography and/or ultrasound[J]. Ann Surg Treat Res, 2016, 90(4): 194−200. DOI: 10.4174/astr.2016.90.4.194 [4] Goldsmith SJ, Parsons W, Guiberteau MJ, et al. SNM practice guideline for breast scintigraphy with breast-specific gamma-cameras 1.0[J]. J Nucl Med Technol, 2010, 38(4): 219−224. DOI: 10.2967/jnmt.110.082271 [5] Brem RF, Floerke AC, Rapelyea JA, et al. Breast-specific gamma imaging as an adjunct imaging modality for the diagnosis of breast cancer[J]. Radiology, 2008, 247(3): 651−657. DOI: 10.1148/radiol.2473061678 [6] Meissnitzer T, Seymer A, Keinrath P, et al. Added value of semi-quantitative breast-specific gamma imaging in the work-up of suspicious breast lesions compared to mammography, ultrasound and 3-T MRI[J]. Br J Radiol, 2015, 88(1051): 20150147. DOI: 10.1259/bjr.20150147 [7] Kuhn KJ, Rapelyea JA, Torrente J, et al. Comparative Diagnostic Utility of Low-Dose Breast-Specific Gamma Imaging to Current Clinical Standard[J]. Breast J, 2016, 22(2): 180−188. DOI: 10.1111/tbj.12550 [8] Tan H, Jiang L, Gu Y, et al. Visual and semi-quantitative analyses of dual-phase breast-specific gamma imaging with Tc-99m-sestamibi in detecting primary breast cancer[J]. Ann Nucl Med, 2014, 28(1): 17−24. DOI: 10.1007/s12149−013−0776−7 [9] Park JS, Lee AY, Jung KP, et al. Diagnostic Performance of Breast-Specific Gamma Imaging (BSGI) for Breast Cancer: Usefulness of Dual-Phase Imaging with 99mTc-sestamibi[J]. Nucl Med Mol Imaging, 2013, 47(1): 18−26. DOI: 10.1007/s13139−012−0176−2 [10] Rechtman LR, Lenihan MJ, Lieberman JH, et al. Breast-specific gamma imaging for the detection of breast cancer in dense versus nondense breasts[J]. AJR Am J Roentgenol, 2014, 202(2): 293−298. DOI: 10.2214/AJR.13.11585 [11] Del VS, Salvatore M. 99mTc-MIBI in the evaluation of breast cancer biology[J]. Eur J Nucl Med Mol Imaging, 2004, 31 Suppl 1: S88−96. DOI: 10.1007/s00259−004−1530−0 [12] 杜勇, 任长征, 龙艳. 乳腺肿块99mTc-MIBI显像规律与其组织病理学关系的初步探讨[J]. 中华核医学杂志, 1997, 17(1): 43−45. DOI: 10.1007/BF02951625
Du Y, Ren CZ, Long Y. The relationship between the 99mTc-MIBI imaging characteristics and the histopathologic features of breast tumors[J]. Chin J Nucl Med, 1997, 17(1): 43−45. DOI: 10.1007/BF02951625[13] Tadwalkar RV, Rapelyea JA, Torrente J, et al. Breast-specific gamma imaging as an adjunct modality for the diagnosis of invasive breast cancer with correlation to tumour size and grade[J]. Br J Radiol, 2012, 85(1014): e212−216. DOI: 10.1259/bjr/34392802 [14] Moscoso A, Ruibal Á, Domínguez-Prado I, et al. Texture analysis of high-resolution dedicated breast 18F-FDG PET images correlates with immunohistochemical factors and subtype of breast cancer[J]. Eur J Nucl Med Mol Imaging, 2018, 45(2): 196−206. DOI: 10.1007/s00259−017−3830−1 [15] Jones EA, Phan TD, Blanchard DA, et al. Breast-specific gamma-imaging: molecular imaging of the breast using 99mTc-sestamibi and a small-field-of-view gamma-camera[J]. J Nucl Med Technol, 2009, 37(4): 201−205. DOI: 10.2967/jnmt.109.063537