腮腺疾病的影像学研究

【作者】 张丹；

【导师】 罗天友；

【作者基本信息】 重庆医科大学 ， 影像医学与核医学， 2019， 博士

【摘要】 第一部分CT纹理分析在腮腺常见上皮源性肿瘤鉴别诊断中的初步研究目的探讨基于CT图像的3D纹理分析技术在鉴别腮腺常见良恶性上皮源性肿瘤中的诊断价值。方法回顾性分析138例经病理证实患有腮腺常见良恶性上皮源性肿瘤患者的CT增强动脉期图像资料。选取动脉期图像导入开源软件ITK-SNAP逐层勾画出肿瘤的边界,得到三维可切割图像,再使用MATLAB2017b软件进行纹理特征提取,获取整个瘤体的纹理参数特征。两两分析、比较不同病理类型腮腺肿瘤相关纹理参数值:平均值(mean)、均匀度(uniformity)、能量(energy)、熵(entropy)、偏度(skewness)、峰度(kurtosis)、对比度(contrast)、相关性(correlation)、熵和(sum entropy)、熵差(difference entropy)之间的差异,P<0.05的纹理参数差异具有统计学意义。利用频率变量绘制并计算受试者工作特征(receiver operating characteristic,ROC)曲线及曲线下面积(area under the curve,AUC),作为肿瘤的特异性、准确性及敏感性的评价标准。结果(1)腺淋巴瘤与上皮源性恶性肿瘤组间熵和、熵、平均值、对比度、熵差、能量差异具有统计学意义,且腺淋巴瘤组熵和、熵、平均值、对比度、熵差、能量均高于上皮源性恶性肿瘤组(P<0.05);(2)多形性腺瘤与腺淋巴瘤组间平均值、均匀度、能量、熵、相关性、对比度、熵和、熵差差异具有统计学意义,且腺淋巴瘤组平均值、能量、熵、相关性、对比度、熵和、熵差高于多形性腺瘤组,而多形性腺瘤组均匀度高于腺淋巴瘤组(P<0.05);(3)上皮源性恶性肿瘤与多形性腺瘤组间平均值、均匀度、能量、熵、相关性、熵和差异具有统计学意义,且上皮源性恶性肿瘤组平均值、能量、熵、相关性、熵和高于多形性腺瘤组,而多形性腺瘤组均匀度高于上皮源性恶性肿瘤组(P<0.05)。(4)腮腺腺淋巴瘤与上皮源性恶性肿瘤组平均值的诊断效能最佳,曲线下面积为0.781,特异度为71.1%,灵敏度为73.8%,最佳阈值为80.88;(5)腮腺腺淋巴瘤与多形性腺瘤组能量的诊断效能最佳,曲线下面积为0.910,特异度为96.1%,灵敏度为73.8%,最佳阈值为42558500;(6)腮腺多形性腺瘤与上皮源性恶性肿瘤组能量的诊断效能最佳,曲线下面积为0.742,特异度为86.3%,灵敏度为55.6%,最佳阈值为31563064。结论(1)基于CT图像的3D纹理分析定量参数熵和、熵、平均值、对比度、熵差、能量能够有效鉴别腮腺腺淋巴瘤与上皮源性恶性肿瘤;(2)平均值、均匀度、能量、熵、相关性、对比度、熵和、熵差能够有效鉴别腮腺多形性腺瘤与腺淋巴瘤;(3)平均值、均匀度、能量、熵、相关性、熵和能够有效鉴别腮腺上皮源性恶性肿瘤与多形性腺瘤。(4)根据联合AUC、敏感度、特异度综合分析得出,平均值是鉴别腮腺腺淋巴瘤与上皮源性恶性肿瘤较为理想的纹理参数;(5)能量是鉴别腮腺腺淋巴瘤与多形性腺瘤较为理想的纹理参数;(6)同时,能量也是鉴别腮腺腺淋巴瘤与上皮源性恶性肿瘤相对较为理想的纹理参数。第二部分腮腺淋巴结结核的影像学表现研究目的探讨腮腺淋巴结结核的临床表现、CT和磁共振成像(magnetic resonance imaging,MRI)影像表现特征。方法回顾性分析25例腮腺淋巴结结核患者的临床资料如患者的性别、年龄、临床症状、结核接触史、临床误诊率、相关辅助检查超声、胸片、PPD(purified protein derivative)及CT和/或MR图像,观察病变的位置、形态、大小、数目、强化方式、邻近腮腺筋膜及皮肤的变化情况、颈部肿大淋巴结的分布情况等影像学表现,由两名经验丰富的放射科医师共同分析。结果该组腮腺淋巴结结核患者中年轻人(年龄<50岁)占72%。80%的患者无明显临床症状,亦无确切结核病暴露史。临床误诊率高达64%,影像学误诊率达60%,其中CT和MRI误诊率分别为71.4%和42.9%。本组病例共发现43个病灶,38(88.4%)个病灶累及浅叶,14例(56%)为多发病灶。结核病灶同侧腮腺筋膜改变有4种类型:局部增厚(40%,n=10);局部破溃伴邻近皮肤增厚(28%,n=7);局部隆起(20%,n=5);没有改变(12%,n=3)。25例病例中有14例(56%)出现颈部淋巴结病,6例(24%)为双侧淋巴结病,淋巴结最常受累的区域为Ⅱb区(n=11),Ⅱa区(n=9)和Ⅲ区(n=7)。在CT图像上病灶有4种强化方式:均匀强化(37.1%,n=13)、不规则囊性强化(37.1%,n=13)、厚壁环形强化(14.2%,n=5)和花环状强化(11.4%,n=4)。MRI上病灶信号:T1加权成像(T1-weighted imaging,T1WI)呈等信号,T2加权成像(T2-weighted imaging,T2WI)呈高信号,扩散加权成像(diffusion-weighted imaging,DWI)呈明显高信号,表观弥散系数(apparent diffusion coefficient,ADC)图呈低信号。结论腮腺淋巴结结核在临床上和影像上易误诊为腮腺肿瘤。腮腺淋巴结结核好发的位置为腮腺浅叶,好发的人群为PPD皮肤试验阳性及伴有颈部淋巴结病的年轻人,CT及MRI图像上病灶同侧腮腺筋膜及病灶周围腮腺实质的水肿有助于腮腺淋巴结结核的诊断及鉴别诊断。更多还原

【Abstract】 PART 1 A PRELIMINARY STUDY OF CT TEXTURE ANALYSIS IN THE DIFFERENTIAL DIAGNOSIS OF COMMON PAROTID EPITHELIAL TUMORSObjective To explore the diagnostic efficacy of 3D texture analysis based on CT image in differentiating common benign and malignant epidermal tumors of the parotid gland.Methods CT enhanced images of 138 patients with common benign and malignant epidermal tumors of parotid gland confirmed by pathology were retrospectively analyzed.Texture features,like mean,Uniformity,Energy,Entropy,Correlation,Contrast,Sum Entropy,Difference Entropy,Skewness and Correlation,were extracted by gray-level Histogram and Gray Level Co-occurrence Matrix.Kolmogorov-Smirnov test or Levene test was used for evaluating the significance of the difference between the benign and malignant parotid tumors on quantitative texture features.Receiver operating characteristic(ROC)curve,based on gold standard of pathology,was draw with prediction frequency.The specificity and sensitivity of that were respectively discussed by the maximum Youden’s index.Results(1)The Difference in Sum Entropy,Entropy,Mean,Contrast,Difference Entropy and Energy between parotid adenolymphoma andepidermal malignancies was statistically significant,and The Sum Entropy,Entropy,Mean,Contrast,Difference Entropy and Energy in the parotid adenolymphoma group were higher than that in the epidermal malignancies group(P<0.05).(2)Mean,Uniformity,Energy,Entropy,Correlation,Contrast,Sum Entropy,Difference Entropy between parotid adenolymphoma and pleomorphic adenoma was statistically significant,and Mean,Energy,Entropy,Correlation,Contrast,Sum Entropy,Difference Entropy in the parotid adenolymphoma group were higher than that in the pleomorphic adenoma group,while Uniformity in the pleomorphic adenoma group was much higher than that in the adenolymphoma group(P<0.05).(3)Mean,Uniformity,Energy,Entropy,Correlation,Sum Entropy between parotid adenolymphoma and epidermal malignancies was statistically significant,and Mean,Energy,Entropy,Correlation,Sum Entropy in the epidermal malignancies group were higher than that in the pleomorphic adenoma group,while Uniformity in the pleomorphic adenoma group was much higher than that in the epidermal malignancies group(P<0.05).(4)The diagnosis efficiency of Mean in the group of parotid adenolymphoma and epidermal malignant tumor was the best,and the AUC was 0.781.The sensitivity was 73.8%,the specificity was 71.1%,and the optimum diagnostic performance was 80.88.(5)The diagnosis efficiency of Energy in the group of parotid adenolymphoma and pleomorphic adenoma was the best,and the AUC was 0.910.The sensitivity was 73.8%,the specificity was 96.1%,and the optimumdiagnostic performance was 42558500.(6)The diagnosis efficiency of Energy in the group of parotid epidermal malignancy and pleomorphic adenoma was the best,and the AUC was 0.742.The sensitivity was 55.6%,the specificity was 86.3%,and the optimum diagnostic performance was31563064.Conclusion Quantitative parameters of 3D texture analysis based on CT images: Sum Entropy,Entropy,Mean,Contrast,Difference Entropy and Energy can effectively distinguish between parotid adenolymphoma and epidermal malignancies.Mean,Uniformity,Energy,Entropy,Correlation,Contrast,Sum Entropy,Difference Entropy can effectively distinguish the pleomorphic adenoma from adenolymphoma of the parotid gland.Mean,Uniformity,Energy,Entropy,Correlation,Sum Entropy can effectively distinguish between parotid pleomorphic adenoma and epidermal malignancies.According to the combined analysis of AUC,sensitivity and specificity,Mean is the ideal texture parameter for distinguishing parotid adenolymphoma from epidermal malignancy.Energy is the ideal texture parameter for distinguishing parotid adenolymphoma from pleomorphic adenoma.Energy is also the ideal texture parameter for distinguishing parotid adenolymphoma from epidermal malignancy.PART 2 TUBERCULOSIS OF THE PAROTID LYMPH NODES: IMAGING FEATURES Objectives To characterize clinical,CT and MRI features of tuberculosis(TB)of the parotid nodes.Methods 25 patients with parotid lymph node tuberculosis were retrospectively analyzed by two experienced radiologists in consensus,including the clinical data such as patients’ age,gender,clinical symptoms,tuberculosis exposure history,clinical misdiagnosis rate and related auxiliary examination ultrasonic,chest radiograph,PPD(purified protein derivative)and CT images and/or MR,observing the location of the lesions,shape,size,number,reinforcement way,the changes of parotid fascia and skin,the distribution of swollen lymph nodes and other imaging findings.Results The youngers(age<50y)account for 72% patients.80%patients were asymptomatic,and with no history of TB exposure.The clinic and imaging misdiagnosed 64% and 60% patients as tumors respectively,A total of 43 lesions were identified.38(88.4%)lesions involved superficial lobe.14(56%)cases had multiple lesions.There were four types of changes of the parotid fascia: local thickening(40%,n=10);local rupture with Thickened adjacent skin(28%,n=7);focal bulge(20%,n=5);and no changes(12%,n=3).Cervical lymphadenopathy was seen in14 out of 25 cases(56%).The lesions were enhanced in four patterns on CT images: homogeneous enhancement(37.1%,n=13),irregular cystic enhancement(37.1%,n=13),thick-walled ring enhancement(14.2%,n=5)and garland-like enhancement(11.4%,n=4).On MRI,the signal of lesions was isointense on T1-weighted image(T1WI),hyperintense on T2-weighted image(T2WI),markedly hyperintense on diffusion-weighted imaging(DWI)and a low signal on the apparent diffused coefficient(ADC)map.Conclusions Tuberculosis of the parotid nodes tend to simulate tumors clinically and radiologically.Their preferential sites are the superficial lobe.The youngers with PPD skin test positive and lesions accompanying with cervical lymphadenopathy,changes of the parotid fascia and parotid parenchyma edema adjacent to the lesions on CT and MRI may be helpful in the diagnosis and facilitate the differential diagnosis.更多还原