基于磁共振成像影像组学的肿瘤内异质性评分模型在卵巢肿瘤恶性风险评估中的应用价值OA

Objective:To calculate an intratumoral heterogeneity(ITH)score using a magnetic resonance imaging(MRI)-based radiomics approach,to construct a preoperative malignancy risk assessment model for ovarian neoplasms by integrating clinical variables and conventional MRI features,and to evaluate its adjunctive value for risk stratification of Ovarian-Adnexal Reporting and Data System(O-RADS)MRI category 4 lesions and for diagnosis by junior physicians.Materials and Methods:This retrospective study included 120 patients with pathologically confirmed ovarian neoplasms,including 52 benign and 68 malignant tumors,the latter including borderline tumors.The patients were randomly assigned to a training set(n=84)and a test set(n=36)in a 7∶3 ratio.After preprocessing T2-weighted images with spectral attenuated inversion recovery fat suppression,intratumoral and 3-mm peritumoral regions of interest were delineated.Radiomics features were extracted using a sliding-window approach,and K-means clustering was used to calculate the ITH score and the combined ITH score.Clinical features,conventional MRI features,and ITH scores were integrated.After screening by univariable analysis and least absolute shrinkage and selection operator regression,independent predictors were identified by multivariable logistic regression to construct a clinical model,an intratumoral model,and a combined model,respectively.All lesions were scored according to O-RADS MRI,and O-RADS MRI category 4 lesions were further analyzed for risk stratification;the adjunctive value of the model for diagnosis by junior physicians was also evaluated.Results:The ITH score showed good intra-and interobserver agreement,with intra-class correlation coefficients of 0.86 and 0.84,respectively.Among the three models,the combined model based on the risk of ovarian malignancy algorithm(ROMA)index,tumor composition,and the combined ITH score achieved the highest diagnostic performance.In the test set,the areas under the receiver operating characteristic curve were 0.805[95％confidence interval(CT):0.662 to 0.948],0.867(95％CI:0.750 to 0.984),and 0.923(95％CT:0.827 to 0.994)for the clinical,intratumoral,and combined models,respectively.The overall diagnostic performance of the O-RADS MRI score was comparable to that of the combined model.In the subgroup analysis of category 4 lesions,the combined model showed potential for further risk stratification,with relatively high diagnostic accuracy and specificity.Conclusion:The combined model based on the ROMA index,tumor composition,and the combined ITH score showed the highest diagnostic performance for preoperative malignancy risk assessment of ovarian neoplasms.It also has the potential to stratify O-RADS MRI category 4 lesions and may assist clinicians in diagnosis.