Study On The Methods Of Image Dynamic Guided Tumor Radiotherapy And Response Evaluation

Radiation therapy is one of the primary treatment modalities for cancer,covering about 70% of cancer patients.The localized treatment modality of radiotherapy makes it closely related to medical imaging.Medical imaging is currently used to achieve radiotherapy image guidance,plan radiotherapy,and assess efficacy.Improving radiotherapy image guidance accuracy and radiotherapy efficacy assessment methods are essential directions for developing precision radiotherapy.Cone-beam CT(CBCT)is one of the most commonly used radiotherapy image guidance modalities.However,CBCT has a long scanning time and is easily affected by respiratory motion,making it challenging to achieve dynamic guidance.Traditional 4-dimensional(4D)CBCT imaging technology is limited by hardware requirements and existing imaging algorithms in the clinical setting,which is difficult to be further promoted.Current efficacy assessment in radiotherapy is based chiefly on post-radiotherapy follow-up examinations,which are time-consuming.The cone-beam CT images scanned in radiotherapy can reflect the tumor changes in radiotherapy,which could be used as a tool for early efficacy assessment in the clinic.The diagnostic value of CBCT images has not been fully explored.Radiomics technology is expected to be used to construct a cone-beam CT-based efficacy assessment model to achieve early efficacy assessment.However,the CBCT images are commonly of poor image quality,which require a rigorous feature screening process to select robust feature.The diagnostic value of CBCT images and its performance in efficacy assessment need to be further investigated.This thesis proposes a 4D CBCT imaging technique in conventional scanning mode,and a CBCT radiomics feature screening method,and a CBCT radiomics-based radiotherapy efficacy assessment method to address the above issues.The 4D CBCT imaging method in standard scan mode is designed to achieve respiratory-related 4D CBCT imaging without changing the existing CBCT scanning equipment and protocol in radiotherapy.For respiratory signal extraction under conventional scanning,we use the position of the diaphragm in CBCT projection as a marker of respiratory motion to achieve respiratory-related projection grouping.To address the problem of severe sparse artifacts in the images reconstructed by phase-sorted projections,we introduced a generative adversarial network to depress the streaking artifacts.However,the training process of the neural network lacks the constraints of the actual physical model and is difficult to be directly oriented to clinical applications.Therefore,we use the corrected images to extract the deformation vector fields between different respiratory phases,and based on this,we perform iterative reconstruction with motion compensation to obtain the final high-quality 4D CBCT images and improve the motion capture capability of CBCT in radiotherapy.The realization of precision radiotherapy also requires personalized treatment strategies for different patients based on medical imaging.For the CBCT radiomics study,we investigated a simulation method for tumor segmentation uncertainties and the effects of slice thickness variation on the stability of CT radiomics features.We proposed a CBCT radiomics feature screening method based on the tumor segmentation stability analysis,volume redundancy analysis,and planning CT interchangeability analysis.Based on the screened CBCT radiomics features,we investigated the value of conventional CBCT features and the variation of CBCT features between different fractions to assess SBRT efficacy in liver cancer.We found that the screened CBCT features had some predictive value for efficacy assessment.The change of CBCT radiomics features between different treatment fractions had a higher efficacy assessment value than the CBCT features in a single fraction.In summary,a 4D CBCT radiotherapy guidance method under conventional CBCT scan and CBCT-based radiotherapy efficacy assessment method are investigated in this thesis.The proposed methods will help improve the accuracy of image-guided radiotherapy,achieve early radiotherapy efficacy assessment,and promote the development of precision radiotherapy.