Differential Diagnosis And Assessment Of Malignancy For Papillary Thyroid Carcinoma Based On MRI Texture Analysis

Objective: The purpose of this study was to investigate the value of preoperative MRI texture analysis in the differential diagnosis of benign thyroid nodules and papillary thyroid carcinoma(PTC),the feasibility of preoperative application of MRI texture analysis in the prediction of cervical lymph node metastasis of PTC,and the feasibility of preoperative application of MRI texture analysis in the prediction of Extrathyroidal extension(ETE).Methods: From January 2015 to January 2019,86 patients with thyroid tumors diagnosed by operation and pathology were collected.The imaging and clinical data of all patients were collected.According to the gold standard of surgical pathology,the patients were divided into benign thyroid nodule group and PTC group,and then PTC was divided into two subgroups,one was the cervical lymph node metastasis group and the non-metastatic group,and the other was t he ETE group and the non-ETE group.All cases were examined by ultrasonography and thyroid MRI before operation.The ROI was manually drawn on the axial images of T1 WI and T2 WI with the largest cross section of the nodule.Texture analysis software was used to extract entropy,standard deviation,contrast,skewness,kurtosis,deficit moment,angular second moment,mean,correlation and other texture parameters.The independent sample t-test(normal distribution with homogeneity of variance)or Mann-Whitney U test(non-normal distribution with homogeneity of variance)was used to compare the texture parameters between the metastatic group and the non-metastatic group.Differences were assessed with the chi-square test for categorical variables.The ROC curve was used to evaluate the texture parameters with statistical significance,and to predict the texture parameters to distinguish benign thyroid nodules from PTC and to evaluate the diagnostic efficacy of PTC malignancy.P Values of less than 0.05 were considered to be statistically significant.Results: 1.The entropy,mean and contrast of benign nodule group in T2 WI image were higher than those of PTC group,while the correlation of PTC group was higher than that of benign nodule group,the differences were statistically significant(P <0.05).There were no significant differences in the standard deviation,angular second moments,skewness,kurtosis,inverse difference moment,and all texture parameters of T1 WI images between benign nodules and PTC(P> 0.05).Among them,the area under the curve of entropy and mean is the largest,which is 0.858 and 0.869,respectively.The best performance is to use entropy value of 6.14 as the threshold to identify PTC,with 87.5% corresponding sensitivity and 76.3% specificity respectively.The best performance is to use mean of 78.15 as the threshold to identify PTC,with 84.8% corresponding sensitivity and 79.2% specificity respectively.The sensitivity of low-level ultrasound physicians when diagnosing PTC was 74.07%(40/54),and the specificity was 62.50%(20/32).The sensitivity of senior ultrasound diagnosis physician is 83.33%(45/54),and specificity is 84.37%(27/32).2.The T2 WI image entropy,correlation,angular second moment,and standard deviation between the PTC cervical lymph node metastasis group and the non-lymph node metastasis group are statistically significant(P <0.05).The area under the ROC curve of entropy was 0.88 4,the threshold was 5.59,the sensitivity was 97.1%,and the specificity was 70%.There was no significant difference in contrast,skewness,kurtosis,inverse moment,mean and all texture parameters of T1 WI images(P> 0.05).The sensitivity and specificit y of radiologists in the diagnosis of cervical lymph node metastasis were 64.7%(22 / 34)and 60.0%(12 / 20).3.The entropy,standard deviation,angular second moment,and correlation based on T2 WI showed significant difference between PTC with and without ETE(P<0.05),among which entropy was the best diagnostic efficiency,the area under ROC curve was 0.865,the diagnostic threshold was 5.82,the diagnostic sensitivity and specificity were 89.7% and 76%,respectively.The mean,skewness,kurtosis,inverse difference moment,contrast,and all texture parameters extracted from T1 WI images were not statistically significant(P> 0.05).The sensitivity and specificity of imaging physicians in diagnosing PTC with ETE were 62.06%(18/29)and 76.00%(19/25),resp ectively.Conclusion: 1.The texture analysis of T2 WI images of conventional MRI sequence is helpful to distinguish benign thyroid nodules from PTC,and its diagnostic efficiency is similar to that of senior ultrasound doctors,but significantly better than that of junior ultrasound doctors.2.The texture analysis of T2 WI image of MRI has certain predictive significance for cervical lymph node metastasis of PTC.3.MRI T2 WI image texture analysis can be demonstrated an effective tool for evaluating the aggressiveness with the histological feature of ETE in PTC.