| Background:In the era of intensity-modulated radiotherapy,nasopharyngeal carcinoma has achieved excellent local control.However,5%-10% of the patients still develop local recurrence after radical radiotherapy.Early diagnosis can create a chance for them to obtain salvage treatment.Radiotherapy can cause soft tissue injuries which contain blood vessels and connective tissue,resulting in morphological and structural changes of soft tissue in the irradiation field,including inflammation,edema,scar retraction,fibrosis and other post-radiotherapy changes,which is challenging to discriminate from local recurrence.Objective:1.Determine the true proportion of local recurrence versus post-radiotherapy changes in patients with enlarging nasopharyngeal mucosal lesion who had completed radical therapy,then explore the factors between them.2.Develop a nomogram based radiomics signature and clinical factors for discriminating tumor recurrence from post-radiotherapy changes in Nasopharyngeal carcinoma.Methods:Part1:This study reviewed patients receiving radical radiotherapy for nonmetastatic nasopharyngeal carcinoma between 2011-2020 who underwent a biopsy for suspicion of post-radiotherapy changes versus local recurrence on MRI imaging.Data including demographics,post-biopsy pathology,magnetic resonance imaging findings was collected.According to the pathological results of biopsy,the collected patients were divided into two groups:pathological biopsy showed recurrence enrolled in local recurrence group and the other went to post-radiotherapy changes group.Fisher’s exact test or the chi-square test was conducted to assess the association between lesion biopsy results and variables of interest.The stratified chi-square test can be an optional for subgroup analysis.Overall survival was determined by Kaplan-Meier methods.Part2:Fifty-one patients with Nasopharyngeal carcinoma suspected of recurrence were included(training cohort=34,validation cohort=17).The imaging features were extracted from the region of interest of all the patients with Matlab2016b software.Ttest and Max-relevance min-redundancy were used to screen,reduce dimension and select features,then construct the radiomics score by support vector machine while a diagnostic nomogram which combined with clinical features for distinguishing tumor recurrence from post-radiotherapy changes was built at the end with logistic regression.Evaluate and validate the predictive performance of the model.Results:Part1:58 patients with biopsied nasopharyngeal mucosa lesions were included.Lesions were biopsied a median of 23.5 months after radical radiotherapy.The local recurrence of nasopharyngeal carcinoma was confirmed in 40 patients(68.97%),and post-radiotherapy changes occurred in 18 patients(31.03%).The MRI clinical diagnoses of 10 patients were corrected after biopsy,and the 17 patients with uncertain MRI diagnoses were confirmed.Most patients had non-keratinizing undifferentiated carcinoma(n=56,96.6%).The median follow-up time(20 months,95%CI:18.25531.745 months)in the group with local recurrence of pathological biopsy was shorter than that in the group with post-radiotherapy changes(26 months,95%CI:11.44840.552 months).In T3 patients,the time from radiotherapy to biopsy was significantly different between the group with local recurrence of pathological biopsy and the group with post-radiotherapy changes.(OR=20,95%CI:1.850-216.178,P=0.009).11 patients among T3 underwent biopsy more than 23 months after radiotherapy,and the positive rate of biopsy was 90.91%(10/11),whereas 4 in 12(33.33%)patients who received biopsies within 23 months after radiotherapy showed tumor recurrence.For 33 patients(58.9%),clinical management was directed or changed by the biopsy results effectively.Part2:Four features were selected from 20,648 imaging features to make a radiomics signature,then combined with clinical risk factors(such as age,sex,TNMstage,T classification,N classification,time from radiotherapy to biopsy)to develop an individualized diagnosis model.The model showed a good performance of prediction and correction in the training and the validation cohort to discriminate tumor recurrence from post-radiotherapy changes,with the AUCs of 0.774 7(95%CI:0.6150.914)and 0.8182(95%CI:0.5999-1.000),respectively.Conclusion:It is risky relying on magnetic resonance imaging solely to distinguish local recurrence with nasopharyngeal carcinoma from post-radiotherapy changes.An individualized assessment combined with other diagnostic methods is preferred.An enlarging lesion should be suspected as local recurrence in patients with T3 nasopharyngeal carcinoma beyond 23 months post-radiotherapy.Further biopsy can confirm the diagnosis and optimize clinical management decisions.Our developed model based radiomics and clinical characteristic could be used to assist the postradiotherapy individualized diagnosis of tumor recurrence in patients with Nasopharyngeal carcinoma. |
PDF Full Text Request