| Head and neck tumors include neck tumors,otolaryngology tumors,oral and maxillofacial tumors.The pathological types are mainly squamous cell carcinoma.As a serious global health problem,head and neck squamous cell carcinomas has an estimated 550,000 new cases and 300,000 deaths each year.Among them,nasopharyngeal carcinoma is one of the most common head and neck malignant tumors.The incidence and mortality of nasopharyngeal carcinoma in China are higher than the global average level,and the death cases account for about 40%of all the deaths of nasopharyngeal carcinoma in the world.Most of the patients with head and neck squamous cell carcinoma have been diagnosed as local advanced stage,which has high recurrence rate and distant metastasis rate.At present,the main treatment for locally advanced head and neck squamous cell carcinoma is radiotherapy combined with chemotherapy.However,local recurrence has become a key challenge after chemoradiotherapy for head and neck squamous cell carcinoma.Therefore,it is of great significance to predict the therapeutic efficacy of patients before treatment or in the early stage of chemoradiotherapy for optimizing clinical treatment strategies and predicting the prognosis of patients.The current TNM staging system has not been very satisfactory in predicting prognosis and therapeutic efficacy.Several studies have explored the predictive power of molecular biomarkers and other clinical features,but these methods have not been tested in a large population,and some tests are expensive.Radiomics refers to the high-throughput extraction of data features from image images.The gray intensity,shape,structure,edge and texture features of images are automatically extracted and quantified,and then analyzed and modeled for disease diagnosis,prognosis judgment and curative effect prediction.With the progress of medical imaging technology and the update of computer algorithms such as artificial intelligence machine learning,medical imaging is expected to shift from the main artificial qualitative diagnosis to the digital information that can be used for quantitative analysis and mining,so as to better support clinical decision-making.In addition,the microscopic characteristics of tumors,such as cells,genes,proteins and even molecular characteristics,are considered to be reflected in the macroscopic images of tumors.Therefore,the molecular mechanism study of radiomics is expected to make a breakthrough.Magnetic resonance imaging(MRI)is a commonly used imaging examination in the diagnosis of head and neck tumors and plays an important role in the diagnosis and disease staging.However,the existing MRI techniques lack sufficient tissue specificity and have many limitations in early prediction of therapeutic efficacy and prognosis.Amide proton transfer imaging(APT)is a novel protein level imaging technique,which reflects the concentration of endogenous free proteins and polypeptides by detecting the concentration of amide proton.Since first reported,APT has been extensively explored for clinical applications in a variety of tumors,including intracranial tumors,liver tumors,breast tumors and prostate tumors.APT imaging has shown a very promising application value in evaluating the therapeutic effect of glioma and rectal cancer.In head and neck tumors,recent studies have found that APT shows good accuracy in differentiating benign and malignant tumors of head and neck.Intravoxel incoherent motion(IVIM)is an extended model of diffusion weighted image(DWI)that can quantitatively evaluate microscopic translational motion in vivo,which may be helpful for obtaining diffuse MRI signals.In this model,biological tissues contain two different environments,molecular diffusion of water in tissues and micro-perfusion of blood in capillaries.One important advantage of IVIM is that it does not need contrast agents.For some patients with renal insufficiency,IVIM is an alternative to perfusion MRI.Multiple parameters D,D*,F and ADC values can reflect the details of intra-tumor perfusion and diffusion.Considering the above-mentioned advantage of IVIM,more and more researchers pay attention to its diagnostic and prognostic role in HNSCC patients.In this study,we investigated the predictive effects of the structural and functional magnetic resonance imaging(MRI)on the efficacy of radiation and chemotherapy for head and neck cancer,including radiomics study of peritumoral and intratumoral magnetic resonance structural imaging,explored the efficacy of new functional sequence APT in early evaluation of treatment efficacy and related histopathological mechanisms in tumor bearing mice model of nasopharyngeal carcinoma,and meta-analysis of MRI functional sequence IVIM related clinical researches.We hope to find reliable imaging markers to predict the therapeutic effect of head and neck cancer from the structural and functional MRI,so as to provide more meaningful imaging reference information for clinical treatment decision-making.Part 1.Prediction of Response to Chemoradiotherapy in Local Advanced Nasopharyngeal Carcinoma:A Peritumoral and Intratumoral Radiomics Study Based on Pretreatment Structural MRIObjectives:The purpose of this study was to determine whether the combination of peritumoral and intratumoral radiomic features on pretreatment MR images could improve the prediction accuracy of chemoradiotherapy response in patients with local advanced nasopharyngeal carcinoma(NPC).Materials and methods:This retrospective study was approved by local ethical committee and informed consent was waived.126 patients with advanced NPC were included and divided into training cohort and validation cohort(ratio:2:1).Treatment response of these patients was determined based on RECIST criterion(version 1.1).ITK-SNAP software(version 3.6.0)was used for three-dimensional manual segmentation of region-of-interest(ROI)on pre-treatment contrast-enhanced T1 weighted images.Tumor ROIs consisted of intratumoral volume(ROI1)and peritumoral volume(ROI2).We performed radiomic features extraction by using pyradiomics(version 3.0)and a total of 2436 radiomic features were extracted from all ROIs.Top performing radiomic features were selected using Random Forest classifier.Prediction models were constructed using training cohort,and then evaluated in validation cohort using area under the receiver operating characteristic curve(AUC).The clinical characteristics of training and validation cohort were compared by using independent samples t-test or ?2 test,which was conducted using Stata/SE version 12.Packages in Python programming language like "pyradiomics","scikit-learn" and"scipy"were used for features extraction,selection,model building and validation.AUC was used for evaluating the performance of the model.All statistical tests were two-sided,and p-values of<0.05 were considered significant.Results:84 patients were randomly assigned to training cohort;42 patients were randomly assigned to validation cohort.General characteristics of the patients were analyzed,including gender,age,overall stage,T-stage,N-stage,hemoglobin,and platelet counts before treatment.There's no difference of these characteristics was found between the two cohorts.A total of 2436 features were extracted from CE-TIW images,1218 from intratumoral volume ROI(ROI 1)and 1218 from peritumoral volume ROI(ROI 2).The top 10 ranked intratumoral radiomic features were used to build model in training cohort,resulting in a mean AUC of 0.82(95%CI:0.80-0.84)in training cohort.In validation cohort,the model showed a mean AUC of 0.81(95%CI:0.78-0.84),sensitivity of 62.5%(95%CI:25.89%-89.76%),specificity of 94.1%(95%CI:78.94%-98.97%).To evaluate the potential increase in predictive power from incorporating peritumoral region radiomic information,all radiomic features(1218 intratumoral and 1218 peritumoral)were combined and ranked based on their feature importance.Seven intratumoral features and three peritumoral features were used to build the joint model,which showed an AUC of 0.90(95%CI:0.89-0.91)in training cohort.In validation cohort,the model showed a mean AUC of 0.91(95%CI:0.89-0.93),sensitivity of 87.5%(95%CI:46.68%-99.34%),specificity of 88.23%(95%CI:71.61%-96.16%).Conclusion:The combination of peritumoral and intratumoral radiomic features on pretreatment MR images could improve the prediction accuracy,which may contribute to improving treatment decision and predicting clinical outcome in locally advanced NPC patients.Part 2.Early Detection Efficacy of Radiotherapy for Nasopharyngeal Carcinoma Using MRI Functional Sequence—Amide Proton Transfer imaging:A Tumor-Bearing Mouse Model StudyObjective:To study the role of Amide Proton Transfer(APT)imaging in early evaluation of radiotherapy efficacy in nasopharyngeal carcinoma(NPC)bearing mice model,and to explore the relevant histopathological mechanism.Materials and methods:Establishment of NPC nude mice model:Four-week-old BALB/C nude mice were selected.Each nude mouse was inoculated with 5 × 106 HNE1 cells and the volume was 0.1 ml.The right neck of the mice was used as the inoculation point.After alcohol disinfection,the needle was injected about 1cm from the inoculation point.when the needle reached the inoculation point,0.1ml cell suspension was injected.In order to exclude the influence of tumor size,nude mice were randomly divided into three groups:high dose group(20Gy/2 fractions),low-dose group(10Gy/2 fractions)and control group(0Gy).MRI scanning was performed before radiotherapy and on the 3,6 and 9 day after radiotherapy.The scanning sequence included T1 weighted,T2 weighted and APT.The volume and MRI image characteristics of tumor at different time points after radiotherapy were quantitatively analyzed.Histopathological examination:tumor tissues of nude mice were obtained on the 3rd day after radiotherapy(the time point at which APT signal began to decrease significantly in the irradiation group),and hematoxylin-eosin staining(HE staining)and terminal deoxynucleoitidyl transferase mediated nick end labeling(TUNEL)immunofluorescence detection were performed.Results:In the high dose group,the tumor volume increased slightly on the third day(from 290 mm3 before radiotherapy to 312 mm3),but no significant change was found on the sixth day(292 mm3),and began to decrease on the ninth day after radiotherapy(208mm3).In the low dose group,the tumor volume continued to increase slightly on the 3rd,6th and 9th day after radiotherapy(326mm3,345mm3,372mm3).In the control group,the tumor volume continued to increase after radiotherapy,and the growth rate was the fastest on the 9th day(331 mm3,459 mm3,651 mm3).In the high dose group,the mean APT value decreased significantly on the 3rd day(from 3.83%before radiotherapy to 2.41%,P<0.001),continued to decrease on the 6th day(1.80%,P=0.001),and increased significantly on the 9th day(2.20%,P=0.009).In the low dose group,the average APT value decreased slightly on the 3rd day after radiotherapy(from 3.52%before radiotherapy to 3.13%,P=0.109),but did not change significantly on the 6th day(3.05%,P=0.64),and increased slightly on the 9th day(3.38%,P=0.122).In the control group,the average APT value tended to be stable(4.22%,4.36%,4.29%).On the 3rd day after irradiation,HE staining in tumor tissue showed nuclear pyknosis and vacuole like changes in high dose group,showing pathological characteristics of cell necrosis.TUNEL staining showed that the proportion of positive apoptotic cells in high dose group was significantly higher than that in control group.Conclusion:In this study,we found that the signal intensity of APT in the high-dose group decreased significantly on the 3rd day after radiotherapy,while the tumor volume decreased on the 9th day after radiotherapy,suggesting that APT can be used to evaluate the therapeutic effect of nasopharyngeal carcinoma in the early stage.After radiotherapy,it was found that high-dose radiotherapy can cause apoptosis and necrosis in tumor tissue,which may be a potential mechanism for the decrease of APT value.Part 3.Early Detection Treatment Response for Head and Neck Squamous Cell Carcinoma Using MRI Functional Sequence——Intravoxel Incoherent Motion imaging:A Meta-AnalysisObjectives:To evaluate the diagnostic accuracy of intravoxel incoherent motion-MRI(IVIM-MRI)for predicting the treatment response in head and neck squamous cell carcinomas(HNSCC)patients.Materials and methods:A comprehensive literature search was performed in Web of Science,PubMed and EMBASE to identify original articles on the diagnostic performance of I VIM in detecting treatment response in HNSCC patients receiving CRT.The IVIM parameters studied were diffusion coefficient(D),pseudodiffusion coefficient(D*),perfusion fraction(f),and apparent diffusion coefficient(ADC).Patients characteristics;study design;treatment flow and timing,image protocols and the diagnostic results(true positive,false positive,false negative and true negative)were extracted.The Quality Assessment of Diagnostic Accuracy Studies 2(QUADAS-2)was used to assess the quality of included studies.A bivariate random-effects model was used to calculate the pooled sensitivity,specificity,diagnostic odds ratio(DOR)and area under receiver operating characteristic curve(AUROC).Hierarchical summary receiver operating characteristic curve(HSROC)was also used to confirm the result.According to Cochrane Collaboration,heterogeneity was suggested to be assessed graphically in the HSROC plot,by observing the prediction ellipse and the scatter of points.Results:Eight studies with a total of 347 patients(with the age ranging from 13 to 73 years)were included in this meta-analysis.There were seven prospective,one retrospective studies.Among these studies,five studies were involved with nasopharynx cancer(NPC),one was involved with hypopharynx squamous cell carcinoma,one was involved with nasal/sinonasal squamous cell carcinoma and the remaining one was involved with a group of head and neck cancers.There was potential introduction of bias in patient selection,index test,reference standard and research flow.Two studies did not report whether a consecutive or random sample of patients were enrolled.The thresholds in all studies were not prespecified.Two studies did not report whether the reference standard results interpreted without knowledge of the results of the index test.Most studies relied solely on radiological follow-up for confirmation of IVIM findings without histopathological confirmation.The pooled sensitivity,specificity,DOR,and AUC were 76%[95%confidence interval(CI)69-82%],81%(95%CI 70-89%),14(95%CI 7-28),and 0.84(95%CI 0.81-0.87)for pre-treatment D,and 70%(95%CI 58-80%),82%(95%CI 66-92%),11(95%CI 5-27),and 0.83(95%CI 0.80-0.86)for ?D,respectively.In addition,the sensitivities and specificities ranged from 41.7 to 94%and 67 to 100%for pre-treatment f,and from 55.7 to 76.5%and 72.2 to 93.3%for pre-treatment apparent diffusion coefficient,respectively.Conclusions:The diffusion-related coefficients pre-treatment D and AD demonstrated high sensitivity and specificity in predicting early treatment response in HNSCC patients.However,because of the variability in reference test and other limitations of included literature,further investigation is needed before implementing any IVIM strategy into clinical practice. |
PDF Full Text Request