Abstract
Background
The accurate estimation of the Ki-67 index of well-differentiated pancreatic neuroendocrine tumors (PanNET) not only guides curative management but also avoids unnecessary treatment.
Purpose
To investigate whether magnetic resonance imaging (MRI) features could preoperatively discriminate the Ki-67 index of well-differentiated G1/G2 PanNETs and to construct an individualized predictive model through developing a MRI-based nomogram.
Material and Methods
Ninety-nine patients with G1/G2 PanNETs were divided into Ki-67 index <5% (n = 58) and Ki-67 index ≥5% groups (n = 41). Multiparametric MRI features of two groups were assessed via uni- and multivariate logistic regression models. The MRI-based nomogram was then developed based on multivariable logistic analysis; the effectivity of the nomogram was validated according to the C-index, calibrations, and a decision curve analysis.
Results
MRI features, including tumor size (odds risk [OR] = 7.791; P = 0.003) and the ADC values (OR = 0.245, P = 0.018) were significant independent risk factors for the high Ki-67 index (≥5%) of G1/2 PanNETs at multivariable analysis. The performance of the MRI-based nomogram (with a C-index of 0.825) was improved compared with that based on tumor size and ADC alone (with C-index values of 0.709 and 0.716, respectively). The calibration curve of the nomogram exhibited good consistency between evaluated and observed outcomes. The decision curve showed that the nomogram incorporating two MRI features had better clinical utilization than single features.
Conclusion
The preliminary MRI-based nomogram could be used to discriminate the Ki-67 index in well-differentiated G1/2 PanNETs.
Get full access to this article
View all access options for this article.