Barkati Maroie, Simard Dany, Taussky Daniel, Delouya Guila
Department of Radiation Oncology, Centre hospitalier de l'Université de Montréal (CHUM), Hôpital Notre-Dame, Montreal, Québec, Canada.
J Med Imaging Radiat Oncol. 2016 Apr;60(2):255-9. doi: 10.1111/1754-9485.12416. Epub 2015 Nov 16.
We assessed the inter- and intra-observer variability in contouring the prostate bed for radiation therapy planning using MRI compared with computed tomography (CT).
We selected 15 patients with prior radical prostatectomy. All had CT and MRI simulation for planning purposes. Image fusions were done between CT and MRI. Three radiation oncologists with several years of experience in treating prostate cancer contoured the prostate bed first on CT and then on MRI. Before contouring, each radiation oncologist had to review the Radiation Therapy Oncology Group guidelines for postoperative external beam radiotherapy. The agreement between volumes was calculated using the Dice similarity coefficient (DSC). Analysis was done using the Matlab software. The DSC was compared using non-parametric statistical tests.
Contouring on CT alone showed a statistically significant (P = 0.001) higher similarity between observers with a mean DSC of 0.76 (standard deviation ± 0.05) compared with contouring on MRI with a mean of 0.66 (standard deviation ± 0.05). Mean intra-observer variability between CT and MRI was 0.68, 0.75 and 0.78 for the three observers. The clinical target volume was 19-74% larger on CT than on MRI. The intra-observer difference in clinical target volume between CT and MRI was statistically significant in two observers and non-significant in the third one (P = 0.09).
We found less inter-observer variability when contouring on CT than on MRI. Radiation Therapy Oncology Group contouring guidelines are based on anatomical landmarks readily visible on CT. These landmarks are more inter-observer dependent on MRI. Therefore, present contouring guidelines might not be applicable to MRI planning.
我们评估了在放射治疗计划中，与计算机断层扫描（CT）相比，使用磁共振成像（MRI）勾勒前列腺床时观察者间和观察者内的变异性。
我们选择了15例既往接受根治性前列腺切除术的患者。所有患者均进行了CT和MRI模拟以用于计划制定。对CT和MRI进行了图像融合。三位有多年前列腺癌治疗经验的放射肿瘤学家先在CT上然后在MRI上勾勒前列腺床。在勾勒之前，每位放射肿瘤学家都必须查阅放射治疗肿瘤学组关于术后体外放射治疗的指南。使用Dice相似系数（DSC）计算体积之间的一致性。使用Matlab软件进行分析。使用非参数统计检验比较DSC。
仅在CT上进行勾勒时，观察者之间的相似性在统计学上具有显著差异（P = 0.001），平均DSC为0.76（标准差±0.05），而在MRI上进行勾勒时平均为0.66（标准差±0.05）。三位观察者在CT和MRI之间的平均观察者内变异性分别为0.68、0.75和0.78。临床靶体积在CT上比在MRI上大19 - 74%。CT和MRI之间临床靶体积的观察者内差异在两位观察者中具有统计学意义，在第三位观察者中无统计学意义（P = 0.09）。
我们发现，与在MRI上勾勒相比，在CT上勾勒时观察者间变异性更小。放射治疗肿瘤学组的勾勒指南基于CT上易于看到的解剖标志。这些标志在MRI上更依赖于观察者之间的一致性。因此，目前的勾勒指南可能不适用于MRI计划。
