Deep Learning for Automated Contouring of Primary Tumor Volumes by MRI for Nasopharyngeal Carcinoma.

Background Nasopharyngeal carcinoma (NPC) may be cured with radiation therapy. Tumor proximity to critical structures demands accuracy in tumor delineation to avoid toxicities from radiation therapy; however, tumor target contouring for head and neck radiation therapy is labor intensive and highly variable among radiation oncologists. Purpose To construct and validate an artificial intelligence (AI) contouring tool to automate primary gross tumor volume (GTV) contouring in patients with NPC. Materials and Methods In this retrospective study, MRI data sets covering the nasopharynx from 1021 patients (median age, 47 years; 751 male, 270 female) with NPC between September 2016 and September 2017 were collected and divided into training, validation, and testing cohorts of 715, 103, and 203 patients, respectively. GTV contours were delineated for 1021 patients and were defined by consensus of two experts. A three-dimensional convolutional neural network was applied to 818 training and validation MRI data sets to construct the AI tool, which was tested in 203 independent MRI data sets. Next, the AI tool was compared against eight qualified radiation oncologists in a multicenter evaluation by using a random sample of 20 test MRI examinations. The Wilcoxon matched-pairs signed rank test was used to compare the difference of Dice similarity coefficient (DSC) of pre- versus post-AI assistance. Results The AI-generated contours demonstrated a high level of accuracy when compared with ground truth contours at testing in 203 patients (DSC, 0.79; 2.0-mm difference in average surface distance). In multicenter evaluation, AI assistance improved contouring accuracy (five of eight oncologists had a higher median DSC after AI assistance; average median DSC, 0.74 vs 0.78; P < .001), reduced intra- and interobserver variation (by 36.4% and 54.5%, respectively), and reduced contouring time (by 39.4%). Conclusion The AI contouring tool improved primary gross tumor contouring accuracy of nasopharyngeal carcinoma, which could have a positive impact on tumor control and patient survival. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Chang in this issue.

A randomized clinical trial comparing prophylactic upper versus whole-neck irradiation in the treatment of patients with node-negative nasopharyngeal carcinoma.

BACKGROUND: This study sought to compare the clinical outcomes of upper versus whole-neck prophylactic irradiation in the treatment of patients with node-negative nasopharyngeal carcinoma (NPC). METHODS: Between November 2005 and June 2012, 301 patients with node-negative NPC were randomly assigned to receive primary plus prophylactic upper neck irradiation (UNI, 153 patients) or primary plus whole-neck irradiation (WNI, 148 patients). Patients in both groups received irradiation to the primary tumor and the upper neck nodal regions, and patients in the WNI group also received irradiation to the lower neck. The main endpoint of the study was to compare the lower neck control rate between the 2 groups. RESULTS: With a median follow-up period of 39 months (range, 6-84 months), no patient in either group had a cervical node relapse. The overall survival at 3 years was 89.5% (95% confidence interval [CI] = 84.1%-95.0%) in the UNI group and 87.4% (95% CI = 81.4%-93.5%) in the WNI group (hazard ratio [HR] = 0.866, 95% CI = 0.41-1.82; P = .70). The 3-year relapse-free survival rate was 89.8% and 89.3% (95% CI = 84.2%-95.3% and 83.7%-94.8%, HR = 0.914, 95% CI = 0.42-2.00; P = .82), and the 3-year metastasis-free survival rate was 91.7% and 90.9% (95% CI = 87.0%-96.5% and 85.7%-96.1%) for the UNI and WNI groups, respectively (HR = 1.007, 95% CI = 0.44-2.32; P = .99). CONCLUSIONS: Prophylactic upper neck irradiation is sufficient for patients with node-negative NPC.

Risk factors and distribution features of level IB lymph nodes metastasis in nasopharyngeal carcinoma.

OBJECTIVE: The objective of this study is to investigate the risk factors and distribution features for level IB metastasis in nasopharyngeal carcinoma (NPC) and provide clinical evidence for defining the indications and clinical target volume (CTV) of prophylactic level IB irradiation. METHODS: We retrospectively analyzed 798 patients with newly-diagnosed, non-metastatic and histologically confirmed NPC underwent intensity-modulated radiation therapy (IMRT). Two sides of neck in each patient have been analyzed separately. The correlations of level IB metastasis and the clinical risk factors were analyzed with Chi-square test and logistic regression model. The risk score model (RSM) of level IB metastasis was calculated by totaling up the scores of each independent variable. We divided level IB into three areas, including anterolateral space of submandibular glands, medial space of the submandibular glands and submandibular glands. RESULTS: Maximal axial diameter (MAD) of level IIA nodes >20mm or extra capsular spread (ES) of level IIA nodes, anterior half of nasal cavity involvement and submandibular gland involvement/compression were independently significantly risk factors for level IB lymph nodes (LNs) metastasis at diagnosis. Two groups based on RSM were obtained: low risk (total score=0-2.5); high risk (4-8.5). The incidence of IB LNs metastasis at diagnosis of the two groups were 0.9% and 6.3%, respectively (P<0.001). The cervical lymph nodes of level IB were distributed in the anterolateral space of submandibular glands. There was no positive/negative LNs inside or medial space of the submandibular glands. CONCLUSION: Level IB LNs metastasis is associated with MAD of level IIA nodes >20mm or ES of level IIA nodes, anterior half of nasal cavity involvement and submandibular gland involvement/compression in NPC patients. Omission of level IB irradiation may be feasible for patients with low-risk IB LNs metastasis at diagnosis. The submandibular gland should not be included in level IB.

[A matched cohort analysis of three-dimensional conformal radiotherapy versus conventional radiotherapy for primary nasopharyngeal carcinoma].

BACKGROUND & OBJECTIVE: Comparing with two-dimensional conventional radiotherapy (2D CRT), three-dimensional conformal radiotherapy (3D CRT) features a precise tumor target volume positioning and a three-dimensional display of the tumor and surrounding normal tissues, therefore, helps to spare normal tissues in the radiotherapy for nasopharyngeal carcinoma (NPC). This study was to explore whether 3D CRT can increase the locoregional control rate while reduce the occurrence and severity of radiation-related sequelae for primary NPC patients. METHODS: Clinical data of 95 primary NPC patients, treated by 3D CRT from Feb. 2001 to Apr. 2004, were reviewed. These patients were then compared with a cohort of 95 NPC patients treated by 2D CRT from Jan. 2002 to Dec. 2003, who were matched in respect of 92 Fuzhou stage, T stage, N stage, sex and age. The short-term and long-term outcomes and radiation-related sequelae of the two groups were compared. RESULTS: There was no significant differences in the 5-year locoregional control rate (89.7% vs. 90.6%, P=0.783), local control rate (91.8% vs. 92.8%, P=0.758), regional control rate (95.6% vs. 97.8%, P=0.427), and distant metastasis-free survival rate (93.1% vs. 86.4%, P=0.152) between 3D CRT group and 2D CRT group. The 5-year overall and progression-free survival rates were slightly higher in 3D CRT group than in 2D CRT group (91.9% vs. 82.0%, P=0.072; 85.7% vs. 71.0%, P=0.087). Eighty-one patients in 3D CRT group and 70 patients in 2D CRT group who had survived without disease progression were compared in respect of late radiation-related sequelae. The occurrence rates of moderate or severe xerostomia and trismus were significantly lower in 3D CRT group than in 2D CRT group (37.0% vs. 71.4%, P<0.001; 6.2% vs. 28.6%, P<0.001). But there were no differences in functional impairment of hearing, memory and injury of cranial nerves between these two groups. CONCLUSION: Comparing with 2D CRT, 3D CRT can reduce the occurrence of late radiation-related sequelae, such as xerostomia and trismus, for primary NPC, and it is likely to prolong overall and progression-free survival.

[Three-dimensional conformal radiotherapy for primary nasopharyngeal carcinoma and analysis of locoregional recurrence].

BACKGROUND & OBJECTIVE: The efficacy of routine two-dimensional radiotherapy for primary nasopharyngeal carcinoma (NPC) is dissatisfied, while three-dimensional conformal radiotherapy (3-D CRT) can optimize irradiation dose distribution. This study was to explore the efficacy of 3-D CRT on NPC, and investigate dosimetric factors of locoregional failure. METHODS: Records of 87 patients with primary NPC, treated with 3-D CRT between Feb. 2001 and Apr. 2004, were retrospectively reviewed. Dosimetric analysis was performed on the patients with locoregional relapse. RESULTS: Of the 87 patients, 5 (5.7%) had local failure, and 2 (2.3%) had regional failure. The overall 3-year locoregional control rate was 90.2%. The 3-year local control rates for stages T1, T2, T3, and T4 patients were 95.0%, 97.0%, 80.1%, and 100%, respectively. Five cases (71%) were marginal or outside failures of the irradiation fields. The occurrence rate of grade 3-4 late complications of the 62 patients progress-freely survived over 12 months was 9.7%. The 3-year overall survival rate and progress-freely survival rate were 88.2% and 80.3%, respectively. The 3-year survival rates were significantly higher in the patients at stages I, II, and III ('92 Fuzhou staging) than in the patients at stage IVa (100%, 100%, 84.7% vs. 47.5%, P<0.001). CONCLUSIONS: Increased locoregional control rate and reduced occurrence of grade 3-4 late complications of primary NPC could be achieved by 3-D CRT through careful delineation of target volumes and design of irradiation fields. Missing of target volumes is a reason of local relapse.

[Irradiation field design according to locoregional relapse sites after radiotherapy of nasopharyngeal carcinoma].

BACKGROUND & OBJECTIVE: The lateral faciocervical field, lateral facial field, and cerrobent blocks were employed conventionally in radiotherapy of nasopharyngeal carcinoma (NPC) in Cancer Center of Sun Yat-sen University since 1990s. This study was to explore rationality of irradiated area of lateral faciocervical field and lateral facial field through analyzing the data of some patients presenting locoregional recurrence after conventional radiotherapy. METHODS: Clinical data of the patients, treated in 1999 and presented locoregional failure and had reports of CT examinations before and after radiotherapy and after recurrence, were reviewed. The regions of recurrent lesions were analyzed. Dosimetric analysis of recurrent regions was performed for the patients with available data of irradiating fields. If no less than 95% of the volume of local recurrent lesion was within 95% isodose curve, it was defined as in-field failure; 20% to less than 95% of the volume was within 95% isodose curve as marginal failure; less than 20% of the volume was within 95% isodose curve as out-field failure. RESULTS: Clinical data of 40 patients could be analyzed, and dosimetric analysis of recurrent regions was performed for 15 patients. Local recurrences mainly occurred in the nasopharyngeal cavity and the skull base bone. Dosimetric analysis showed 9 cases of in-field failure, 4 cases of marginal failure, and 2 cases of out-field failure. For the 6 cases of marginal and out-field failure, the main reasons of recurrence were low dose area in target volume due to not strictly following the guidelines of defining field limit, or inappropriately estimating the involved regions of lesions. CONCLUSION: Irradiated area of lateral faciocervical field and lateral facial field employed conventionally in radiotherapy of NPC is rational.