Zaghi Soroush, Hendizadeh Leenoy, Hung Tony, Farahvar Salar, Abemayor Elliot, Sepahdari Ali R
Department of Head and Neck Surgery, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA.
Department of Head and Neck Surgery, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA.
Am J Otolaryngol. 2014 Nov-Dec;35(6):713-8. doi: 10.1016/j.amjoto.2014.07.013. Epub 2014 Jul 22.
To validate an MRI algorithm characteristic of pleomorphic adenoma (PA).
Cross-sectional analysis.
Academic tertiary-care medical center.
A radiologic algorithm for the MRI diagnosis of PA was developed on the basis of five "high probability" criteria that all must be fulfilled for the MRI to qualify as a positive test result: bright T2-signal, sharp margins, heterogeneous nodular enhancement, lobulated contours, T2-dark rim. We then identified MRI images from our institutional database to test the diagnostic accuracy of the proposed algorithm.
A total of 103 parotidectomy cases with adequate MRI studies were identified (pleomorphic adenoma n=41, mucoepidermoid carcinoma n=11, Warthin's tumor n=8, adenoid cystic carcinoma n=6, oncocytoma n=6, acinic cell carcinoma n=5, salivary duct carcinoma n=5, and other n=21). Eighteen of 21 cases that met all five "high probability" MRI criteria were consistent with PA on final histopathology; 3 were consistent with carcinoma. MRI had a specificity of 95.1% [95% confidence interval: 85.6-98.7%] and sensitivity of 43.9% [95% C.I.: 28.8-60.1%] for PA. The positive predictive value was 85.7% [95% C.I.: 70.4-100%] and the negative predictive value was 71.9% [95% C. I.: 62.0-81.9%]. The overall diagnostic accuracy was 74.8% [95% C.I.: 66.2-83.3%].
A "high probability" MRI is about 95% specific for pleomorphic adenoma. A subset of patients with MRI imaging that is highly suggestive of PA may reliably avoid further workup. The value of MRI in this setting is especially useful if preoperative fine needle aspiration is not readily available. A significant proportion of PAs, however, have indeterminate imaging features that overlap considerably with other benign and malignant lesions.
验证一种多形性腺瘤（PA）特征性的MRI算法。
横断面分析。
学术性三级医疗中心。
基于五项“高概率”标准开发了一种用于PA的MRI诊断的放射学算法，MRI要获得阳性检测结果必须满足所有这些标准：T2信号高、边缘清晰、不均匀结节状强化、分叶状轮廓、T2低信号边缘。然后我们从机构数据库中识别MRI图像，以测试所提出算法的诊断准确性。
共识别出103例有充分MRI研究的腮腺切除术病例（多形性腺瘤41例，黏液表皮样癌11例，沃辛瘤8例，腺样囊性癌6例，嗜酸细胞瘤6例，腺泡细胞癌5例，涎腺导管癌5例，其他21例）。符合所有五项“高概率”MRI标准的21例病例中，18例最终组织病理学结果与PA一致；3例与癌一致。MRI对PA的特异性为95.1%[95%置信区间：85.6 - 98.7%]，敏感性为43.9%[95%置信区间：28.8 - 60.1%]。阳性预测值为85.7%[95%置信区间：70.4 - 100%]，阴性预测值为71.9%[95%置信区间：62.0 - 81.9%]。总体诊断准确性为74.8%[95%置信区间：66.2 - 83.3%]。
“高概率”MRI对多形性腺瘤的特异性约为95%。一部分MRI成像高度提示PA的患者可能可靠地避免进一步检查。如果术前细针穿刺不容易获得，MRI在这种情况下的价值特别有用。然而，相当一部分PA具有不确定的影像学特征，与其他良性和恶性病变有很大重叠。
