| Part I:Application of radiomics based on multi-parameter magnetic resonance in predicting T-stage of rectal cancerObjective:To explore the value of radiomics feature model based on multi-parameter magnetic resonance T2 WI,enhanced T1 WI and DWI sequence images in predicting Tstage of rectal cancer,so as to provide data support for clinical application of radiomics technology.Methods:The clinical and imaging data of 256 patients with rectal cancer confirmed by operation or biopsy in ou hospital from January 2019 to November 2020 were collected retrospectively.All patients underwent pelvic MRI before operation.The scanning sequence included T2 WI,enhanced T1 WI and diffusion weighted imaging（DWI）.First,all cases were randomly divided into training group and verification group according to the ratio of 7:3.Then,according to the standard of pathological stage,T1-T2 stage was divided into early stage group（n = 154）and T3-T4 stage（n = 102）was divided into local progressive stage group.Then manually sketch the two-dimensional region of interest（ROI）along the edge of the lesion on the image,and extract radiomics features of three sequences.The minimum maximum normalization,optimal feature screening and lasso logistic regression model are used to reduce the dimension of data,screen the best features and construct the radiomics model.Finally,the area under the receiver operating characteristic（ROC）curve（AUC）was used to evaluate the diagnostic efficacy of different models.Results: In the training group and the verification group,the AUC values of the area under the ROC curve of the radiomics model constructed by a single sequence of T1WI+,T2 WI and DWI were 0.84,0.70,0.58 and 0.80,0.78 and 0.80 respectively,but the AUC values of the area under the ROC curve of the combined three sequence models were 0.96 and 0.95 respectively,with the highest efficiency.Conclusion:Radiomics based on multi-parameter MRI can be used as a non-invasive method to diagnose preoperative T-staging of rectal cancer and provide support for clinical decision-making of rectal cancer.Part II:Application of radiomics based on multi-parameter magnetic resonance in pathological grading of rectal adenocarcinomaObjective: To explore the clinical value of radiomics model based on multi-parameter MRI in predicting the pathological grading of rectal adenocarcinoma before operation.Methods: The clinical and imaging data of 153 patients with rectal adenocarcinoma confirmed by operation or biopsy in our hospital were collected retrospectively.All patients underwent pelvic MRI before operation.First,all cases were randomly divided into training group and verification group according to the ratio of 7:3.Then,according to the standard of pathological differentiation,the pathological high differentiation and medium differentiation were low-grade rectal adenocarcinoma（n = 97）and low differentiation was high-grade rectal adenocarcinoma（n = 56）.The rest is the same as part I.Results: In the training group and the validation group,the AUC values of the area under the ROC curve of the joint sequence model were 0.84 and 0.82 respectively,with the highest efficiency.The AUC values of imaging radiomics models constructed by DWI,T2 WI and T1 WI + single sequence in training group and verification group were0.82,0.70,0.60 and 0.75,0.68 and 0.64 respectively,which were lower than those of combined sequence.Conclusion: The prediction model based on multi-sequence MRI radiomics has certain application value for preoperative evaluation of pathological grading of rectal adenocarcinoma.Part III:Application of clinical-radiomics nomogram based on multi-parameter magnetic resonance in predicting lymph node metastasis of rectal cancerObjective: To explore the value of clinical-radiomics nomogram based on multiparameter magnetic resonance in predicting lymph node metastasis of rectal cancer.Methods: 242 patients with rectal cancer who were confirmed by operation or biopsy and underwent MRI before operation were collected in our hospital retrospectively.According to the pathological results,127 cases were positive for lymph node +（LN+）and 115 cases were negative for lymph node-（LN-）.Then manually sketch the two-dimensional region of interest（ROI）along the edge of the lesion on the image,and extract the radiomics features of three sequences.Secondly,the minimum maximum normalization,optimal feature screening and lasso logistic regression model are used to reduce the dimension of data,screen the best features and construct radiomics model.In addition,eight related clinical risk factors were included,and the clinical risk factors screened by multifactor logistic regression were used to construct the clinical characteristic model.Finally,three prediction models were established: clinical model,radiomics model and clinical-radiomics model.The area under the receiver operating characteristic（ROC）curve（AUC）was used to evaluate the diagnostic efficacy of the model and construct a nomogram.Results: In the training group and validation group,the areas under the curve of clinical-radiomics model,radimocs model and clinical model are 0.86,0.84,0.62 and0.77,0.71 and 0.56 respectively.The clinical-radiomics model has the highest diagnostic efficiency.Conclusion: The clinical-radiomics nomogram based on multi-parameter MRI radiomics and clinical features can provide clinical decision support for the judgment of preoperative lymph node metastasis of rectal cancer. |
PDF Full Text Request