Multimodal Magnetic Resonance Imaging-based Precise Solution For Diagnosis And Treatment In Nasopharyngeal Carcinoma

In China,nasopharyngeal carcinoma（NPC）is a common head and neck malignant tumor.Local recurrence and distant metastasis are two major failure modes of NPC.On the one hand,the value of bio-intensity modulated radiation therapy and the normal organ tolerated dose in the era of intensity-modulated radiation therapy need to be re-evaluated.On the other hand,predicting the risk of tumor metastasis based on clinical stage is not sufficient to guide individualized treatment effectively,early prediction of metachronous oligometastases（MO）is extremely challenging.Multimodal magnetic resonance imaging（MRI）has a wide application prospect in the field of target delineation,evaluation of radiation related side effects and prognosis prediction of head and neck tumors.Deep learning radiomics is an important direction for the medico-engineering cooperation,and it has an important application potential in the prediction of tumor prognosis.This dissertation utilizes multimodal MRI to investigate individual dose painting of locally advanced NPC,dose limitation of radiation-induced brainstem necrosis（RIBN）,and intelligent prediction of MO in NPC with the goal of resolving those scientific issues.In the first part of this dissertation,we compared the survival and side effects of patients who were treated with or without multimodal MRI guided individual dose painting after conventional chemoradiotherapy course.The result showed that the median follow-up of patients in the dose painting group and the conventional group were 48months（range 8-88 months）and 52 months（range 6-90 months）respectively.The 5 year overall survival（OS）,distant metastasis-free survival（DMFS）、progression-free survival（PFS）and local recurrence-free survival（LRFS）of patients in the dose painting group and the conventional group were 88%vs 82.5%（p=0.244）,86.1%vs 83.3%（p=0.741）,82.2%vs 76.6%（p=0.286）and 89.1%vs 80.1%（p=0.029）respectively.Multivariate analysis showed that MRI guided individual dose painting was independent prognostic factor for LRFS（HR 0.386,95%CI 0.163-0.909,p=0.03）.Multimodal MRI showed great application value in dose painting of NPC.In the second part of this dissertation,the incidence of brainstem necrosis was evaluated by MRI,and logistic regression was used for screening dose-volume parameter and constructing of logistic dose response model.The incidence of RIBN was 1.25%（6/479）,and the median time to RIBN after treatment was 28.5 months（range 18-48months）.The brainstem dose was higher than that in patients without necrosis.ROC curve showed that the AUC of D_max was the largest（0.987）.Furthermore,logistic stepwise regression revealed that D_max was the most important dose factor.D_max<69.59Gy was suggested to be the dose restriction for brainstem in the era of IMRI.MRI was valuable in evaluating brainstem necrosis and led to better organ protection and precision radiotherapy in combined with dose analysis.In the third part of this dissertation,the MRI,clinical variable（CV）and follow-up data of the186 patients with NPC were collected.Gross tumor volume（GTV）and lymph node gross tumor volume（GTVln）prior to treatment were defined on T1WI,T2WI and CE-T1WI.After image normalization,the deep learning platform Python（version 3.9.12）was used in Ubuntu 20.04.1 LTS to construct automatic tumor detection and MO prediction model.We found that the overall performance of automatic tumor detection model was the model based on CE-T1WI.Automatic segmentation algorithm based on Mask Scoring R-CNN had the best overall performance for automatic identification of tumor and metastatic lymph nodes on CE-T1WI images（m AP@0.5=57.8%）.When the Mask R-CNN instance segmentation algorithm was used for automatic detection,the AUCs of the MO prediction model based on T1WI,T2WI and CE-T1WI were 0.722,0.695 and 0.733,respectively.When Cascade Mask R-CNN and Mask Scoring R-CNN instance segmentation algorithm were used for automatic detection,similar prediction model could be acquired.After adding CV,the prediction ability of the prediction model based on T1WI,T2WI and CE-T1WI was further improved under the three automatic segmentation algorithms.The largest AUC（0.775,95%CI 0.606-0.945,p=0.001）was acquired in the prediction model based on the CE-T1WI and CV when Mask R-CNN automatic segmentation algorithm was used.By comparing the 3-year survival of high-risk and low-risk patients based on the combined model,we found that the 3-year DMFS and OS of high and low MO risk patients were 11.4%vs 95%and 85.3%vs 97%respectively（p<0.05）.Deep learning based on multimodal MRI is expected to accurately predict MO of NPC.Conclusions:（1）Multimodal MRI guided individual dose painting is a viable strategy for improving local control in patients with locally advanced NPC.The side effects of the treatment were tolerable.（2）High dose irritation is linked to brainstem necrosis.D_max is the most important predictive dosimetric factor for RIBN.D_max of brainstem is suggested to be the dose limitation parameter.We propose that the limitation dose for brainstem was D_max<69.59Gy.（3）The intelligent prediction model based on magnetic resonance imaging alone or in combination with clinical data has excellent performance in automatic tumor detection and MO prediction for NPC patients,and is worthy of clinical application.The main innovations of this study are as follows:1.A method for individual dose painting of NPC based on multimodal MRI was proposed;2.Recommendation for the dose restriction of brainstem necrosis in the era of IMRT was provided;3.A prediction method for MO of NPC based on multimodal MRI and deep learning was established.