The Application Of MRI Texture Analysis In The Grading Of Pancreatic Neuroendocrine Neoplasms And Differentiation Non-functional Pancreatic Neuroendocrine Neoplasms From Solid Pseudopapillary Neoplasms Of The Pancreas

Objective:To evaluate the value of MRI texture analysis for identifying non-functional pancreatic neuroendocrine neoplasms?NF-PNENs?and solid pseudopapillary neoplasms?SPNs?.Methods:This retrospective study included 119patients with NF-PNENs and SPNs diagnosed by pathology.Texture analysis was performed using MaZda software.Patients were divided into training sample?101cases?and validation sample?18 cases?.Raw data analysis?RDA?,principal component analysis?PCA?,linear discriminant analysis?LDA?and nonlinear discriminant analysis?NDA?were used to classify NF-PNENs and SPNs in the training sample.NDA with artificial neural network classifier was used for validation sample.The results were reported as misclassification rates.The complete MR images were evaluated by an experienced senior radiologist without knowledge of the MaZda software'results,pathological results and clinic information to differentiate the two types of pancreatic neoplasms.The results were also reported as misclassification rates.In addition,30 texture features selected by MaZda for each MRI sequence were compared between the two groups.Results:The lowest misclassification rate of identification of NF-PNENs and SPNs was 7.92%?6/101?in the training sample.It was obtained with NDA using FPM method on DWI,but there was no significant difference among the sequences.The misclassification rate of the radiologist?34.65%,35/101?was significantly higher than that of NDA for all sequences.In the training sample,entropy and sum entropy were the optimal texture features on DWI and pre-contrast T₁WI+fs,while the mean and percentile seemed to be the more discriminative features on DCE-T₁WI+fs.The validation results were good in the arterial phase?AP?and delayed phase?DP?.Histogram features showed statistically significant differences on DCE-T₁WI+fs in both the training and validation sample.Conclusion:Texture analysis based on MR images can distinguish between NF-PNENs and SPNs,and histogram features of DCE-T₁WI+fs images had high discriminancy efficiency.Application of texture analysis based on MR images in the grading of pancreatic neuroendocrine neoplasmsObjective:To evaluate the value of MRI texture analysis for identifying pathological grades of pancreatic neuroendocrine neoplasms.Methods:This retrospective study included 151 PNENs patients?84 G1 patients,55 G2 patients,12 G3 patients?diagnosed by pathology.Texture analysis was performed using MaZda software.Patients included were divided into training sample?136 cases?and validation sample?15 cases?.Raw data analysis?RDA?,principal component analysis?PCA?,linear discriminant analysis?LDA?and nonlinear discriminant analysis?NDA?were used to identify three pathological grades in the training sample.The results were reported as misclassification rates.NDA with artificial neural network classifier was used for validation sample.In addition,30 texture features selected by MaZda for each MRI sequence were investigated.Results:The lowest misclassification rate of identifying three pathological grades was19.12%?26/136?in the training sample,but G3 could not be distinguished from G1 and G2.The lowest misclassification rate of identifying of G1 and G2 was 12.70%?16/126?in the training sample.Co-occurrence matrix and run-length matrix features showed statistically significant differences among three pathological grades.Statistically significant features between G1 and G2 were from T₂WI+fs,DWI and ADC map.They were mainly co-occurrence matrix and run-length matrix features.Conclusion:MRI texture analysis is helpful to identify G1 and G2 of PNENs,and co-occurrence matrix and run-length matrix features provided more information.