Diagnostic value of a multiparametric magnetic resonance imaging-based nomogram incorporating apparent diffusion coefficient in preoperative differentiation of phyllodes tumors from breast cancer: a retrospective study

Abstract

Background: Preoperative differentiation between phyllodes tumors (PTs) and breast cancers (BCs) remains a clinical challenge due to their overlapping imaging characteristics. The aim of this study was to evaluate the diagnostic performance of multiparametric magnetic resonance imaging (MRI), incorporating morphological, hemodynamic, and apparent diffusion coefficient (ADC) parameters, to develop a nomogram model for differentiating PTs from BCs, and to preliminarily explore the association between ADC parameters and the histologic grade of PTs. Methods: A total of 92 patients (mean age, 46.04 years) with pathologically confirmed PTs (n=31) or BCs (n=61) who underwent 3.0 Tesla (T) breast MRI at the First Affiliated Hospital of Soochow University between June 2018 and June 2025 were retrospectively enrolled. Multiparametric MRI features, including morphological characteristics, hemodynamic parameters, and quantitative diffusion metrics [mean ADC (ADC mean ) and relative ADC (rADC)], were assessed for all lesions. Intergroup differences were compared using the Kolmogorov-Smirnov test, independent samples t-test, one-way analysis of variance (ANOVA), Mann-Whitney U test or Kruskal-Wallis H test, Chi-squared test, or Fisher’s exact test, as appropriate. A multivariable logistic regression analysis was performed to identify independent predictors and construct a diagnostic nomogram. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curve analysis, decision curve analysis (DCA), and calibration curve analysis. Finally, 5-fold internal cross-validation was used to assess the model’s stability. Results: PTs were characterized predominantly by a lobulated morphology, hyperintensity on T2-weighted imaging (T2WI), and T2 hypointense septations, whereas BCs typically presented as irregular masses with spiculated margins and peritumoral edema. Regarding dynamic contrast-enhanced MRI (DCE-MRI) features, PTs predominantly demonstrated rapid initial enhancement with heterogeneous or dark internal septations, and time-intensity curves (TICs) mainly exhibited Type I/II patterns. PTs demonstrated significantly higher ADC mean , rADC1 (lesion-to-pectoralis muscle ratio), and rADC2 (lesion-to-contralateral gland ratio) values compared with the BCs (all P<0.001). Subgroup analysis revealed a stepwise decrease in ADC mean from benign PTs to borderline/malignant PTs and BCs. The diagnostic nomogram integrating T2WI signal intensity, T2 hypointense septations, TIC pattern, and ADC mean achieved an apparent area under the curve (AUC) of 0.997. Following rigorous internal 5-fold cross-validation with Firth’s penalty to mitigate potential overfitting, the model maintained a robust average AUC of 0.978 [95% confidence interval (CI): 0.951–1.000]. DCA demonstrated the net clinical benefit of the combined model. Conclusions: Multiparametric MRI provides significant diagnostic value in preoperatively differentiating PTs from BCs. Although the subgroup analysis regarding PT histological grading remains preliminary due to limited sample sizes, both ADC mean and rADC offer valuable quantitative insights, with rADC serving as a complementary metric. Crucially, the combined diagnostic nomogram demonstrated robust diagnostic performance despite internal penalization. Pending prospective external validation, this tool may inform individualized diagnosis and optimize surgical decision-making.