MRI Radiomic Model For Discrimination Of Adamantinomatous Craniopharyngioma And Squamous Papillary Craniopharyngioma

PurposeTo explore the value of structural MRI and machine learning-based multiparametric magnetic resonance imaging radiomic model for preoperative discrimination between adamantinomatous craniopharyngioma and squamous papillary craniopharyngioma.Materials and MethodsA total of 164 patients from two medical centers were enrolled in this study.Patients from Nan Fang Hospital were divided into a training cohort(N=99)and an internal validation cohort(N=33).Patients from Yan Ling Hospital were used as the external independent validation cohort(N=32).In the first part,we compared the clinical and structural MRI features between adamantinomatous craniopharyngioma and squamous papillary craniopharyngioma,and constructed a clinicoradiological model to distinguish adamantinomatous craniopharyngioma from squamous papillary craniopharyngioma.In the second part,we construct a machine learning-based multiparametric magnetic resonance imaging(axial T1WI,T2WI,CET1WI)radiomic model.The tumor VOI was delineated with relevant scientific research software,and the radiomic features were extracted.Optimal radiomic feature selection was performed by SelectKBest,the least absolute shrinkage and selection operator algorithm,and support vector machine(SVM)with a recursive feature elimination algorithm.Models based on each sequence or combinations of sequences were built using machine learning algorithm(SVM)and were used to differentiate adamantinomatous craniopharyngioma and squamous papillary craniopharyngioma in the training cohort.The diagnostic performance of these models was validated with an internal validation cohort,and the diagnostic performance of the best model was further validated with an external validation cohort.ResultsIn the first part,No differences were observed between the adamantinomatous craniopharyngioma and squamous papillary craniopharyngioma in terms of gender,shape,composition,or pituitary stalk morphology in the training cohort(P=0.100-0.777).However,age,tumor location,signal intensity of tumor cysts on noncontrast T1WI,and the enhancement pattern differed significantly between the two groups in the training cohort.The clinicoradiological model was developed combining age,tumor location,signal intensity of tumor cysts on noncontrast T1WI,and the enhancement pattern.The AUC values of the clinicoradiological model were 0.677,0.655,and 0.671 in the training cohort,internal and external validation cohorts,respectively.In the second part,the radiomic model with the combination of the three sequences has the highest diagnostic efficiency.Seven texture features,three from T1WI,two from T2WI,and two from CET1WI,were selected and used to construct the radiomic model.The AUC values of the radiomic model were 0.899,0.810,and 0.920 in the training cohort,internal and external validation cohorts,respectively.ConclusionThe machine learning radiomics model based on multiparametric MRI showed better performance than the clinicoradiological model for distinguishing adamantinomatous craniopharyngioma from squamous papillary craniopharyngioma.