引入靶区外扩预测放射治疗中危及器官自动分割的平均剂量偏差

目的 观察引入靶区外扩预测放射治疗(放疗)中自动分割危及器官(OAR)的平均剂量偏差的价值。方法 将100例接受放疗的直肠癌患者随机分为训练集(n=30)和测试集(n=70)。对训练集手动分割CT图中的靶区,之后分别对膀胱、小肠和双侧股骨头4个OAR进行手动和自动分割。根据自动分割的OAR设计放疗计划,得到对应的剂量分布;利用Python程序统计每个OAR与靶区外扩环重叠区域内的剂量平均值,以之作为代表剂量,用于预测测试集手动与自动分割平均剂量的差异,比较预测平均剂量与实际平均剂量的差异。再次随机将100例分为训练集、测试集各50例,重复上述过程。结果 首次预测显示,测试集70例中,69例膀胱预测与实际剂量差异均<0.5 Gy,69例小肠预测与实际剂量差异均<3 Gy,全部70例双侧股骨头预测与实际剂量差异均<0.5 Gy;对于膀胱、小肠和左、右侧股骨头,预测与实际平均剂量差异的一致性相关系数(CCC)分别为0.96、0.86、0.81和0.69。第2次预测显示,测试集50例中,46例膀胱的预测与实际剂量差异均<0.5 Gy,49例小肠的预测和实际剂量差异均<3 Gy,所有病例双侧股骨头的预测和实际剂量差异均<0.5 Gy;对于膀胱、小肠和左、右侧股骨头,预测与实际平均剂量差异的CCC分别为0.97、0.90、0.82和0.78。结论 引入靶区外扩可有效预测直肠癌放疗中自动分割OAR产生的剂量偏差。

Predicting the mean dose deviation of radiotherapy caused by automatic segmentation of organ at risk based on target expansion

Objective To explore the value of predicting the mean dose deviation of radiotherapy caused by automatic segmentation of organ at risk (OAR) based on target expansion. Methods Totally 100 patients with rectal cancer who underwent radiotherapy were enrolled, 30 cases were selected as the training set, and the remaining 70 cases as the test set. For the training set, the targets were manually contoured, while the OAR (bladder, small intestine, left and right femur heads) were defined by both manual and automatic segmentation. Taken the automatic segmented organs as the objective items, the radiotherapy plans were optimized, and the dose distributions based on automatic segmentation were acquired. A Python program was used to calculate the average dose in the overlapping regions of each OAR and the target expansion ring, which was then applied as a representative dose to cases in the test set for predicting differences of the mean doses between manually and automatically delineated ones, and the differences of the predicted and actual mean doses were compared. Again, the patients were randomly divided into training set and test set (each n=50), and the above process was repeated. Results For the first test set (n=70), the difference of predicted and actual dose of bladder was <0.5 Gy in 69 cases, of small intestine was <3 Gy in 69 cases, while of bilateral femoral heads were <0.5 Gy in all 70 cases. The concordance correlation coefficient (CCC) between predicted and actual dose for bladder, small intestine, left and right femoral head was 0.96, 0.86, 0.81 and 0.69, respectively. For the second test set (n=50), the difference of predicted and actual dose of bladder was <0.5 Gy in 46 cases, of small intestine was <3 Gy in 49 cases, while of bilateral femoral heads were <0.5 Gy in all 50 cases. The CCC between predicted and actual dose for bladder, small intestine and left and right femoral head was 0.97, 0.90, 0.82 and 0.78, respectively. Conclusion Target expansion could be used to predict the mean dose deviation due to OAR automatic segmentation in radiotherapy of rectal cancer.