Prediction Of The Texture Of Pituitary Adenoma Using MR Texture Parameters Based On Imageomics Model

Objective:Based on the texture parameters of ce-t1 wi and T2 WI images,the texture prediction model of pituitary macroadenoma was constructed,and the effects of different field intensity and equipment of different suppliers on the prediction efficiency of each model were analyzed.Materials and methods:This study was a retrospective study.180 patients with pituitary macroadenoma diagnosed in the Department of Radiology,China-Japan union hospital of Jilin university from January 2011 to September 2020 were selected as the research objects.The patients without pathological results and patients with incomplete images were excluded,and the remaining 130 patients were included.Masson staining was performed on the pathological specimens of all patients to evaluate the expression of collagen to determine the softness and hardness of the texture.T1 enhanced images and T2 WI images of all patients were collected at the same time,and two radiologists of different years used ITK-SNAP software to manually delineate the lesion layer by layer to generate a region of interest(ROI).Use AK software to extract image features,use regression analysis to reduce the dimensions of the features,and use machine learning methods to establish imaging omics models based on T1 enhancement(CE-T1WI)and T2 WI images for pituitary macroadenoma For texture prediction,the predictive performance of each model is evaluated by receiver operating characteristic curve(ROC),and the predictive performance of scanned images of different inspection equipment is tested.Results:Based on the imaging omics label established based on the eigenvalues ??selected from the T1 enhanced image,the area under the ROC curve(AUC)of the training group and the validation group were0.828 and 0.802,respectively,the accuracy rates is 0.787 and 0.733,and the specificities is 0.779 and 0.779,the sensitivities were 0.794 and 0.800,respectively;the imaging omics label established based on the feature values ??filtered by the T2 WI images,the area under the ROC curve(AUC)of the training group and the validation group were 0.845 and0.791,respectively,and the accuracy is 0.779 and 0.767,specificity is0.750 and 0.767,sensitivity is 0.809 and 0.767,respectively.The area under the ROC curve(AUC)of the CE-T1 WI image combined with T2 WI image model in the training group and the validation group is0.902 and 0.862,respectively,the accuracy rates is 0.816 and 0.833,the specificity is 0.794 and 0.833,and the sensitivity is 0.838 and 0.833.The area under the AUC curve of the predicted performance of the CE-T1 WI image model in 1.5T and 3.0T devices is 0.50 and 0.61,respectively;the area under the AUC curve of the predicted performance of the CE-T2 WI image model in 1.5T and 3.0T devices is 0.76,and 0.76,respectively.0.834;The area under the AUC curve of the combined model’s predicted performance in 1.5T and 3.0T equipment is 0.58 and 0.56,respectively.The area under the AUC curve of the CE-T1 WI image model predicting effectiveness in Siemens and Philips devices is 0.60 and 0.54,respectively;the area under the AUC curve of the T2 WI image model predicting effectiveness in Siemens and Philips devices is 0.83 and 0.84;the joint model T2 WI image The area under the AUC curve of the model’s predicted performance in Siemens and Philips equipment is 0.53 and 0.52,respectively.Conclusion:1.The imaging model based on T1 enhanced image and T2 WI image can predict the texture of pituitary macroadenoma.2.The model based on T2 WI image is less affected by different field strength and different supplier equipment.