Quantitative imaging–pathology correlation in nodular fasciitis and diagnostic discrimination from sarcoma

Objectives To investigate associations between key magnetic resonance imaging features of nodular fasciitis (NF) and quantitative histopathologic biomarkers, and to evaluate whether their integration provides complementary information for differentiating NF from soft tissue sarcoma (STS) in diagnostically equivocal settings. Materials and methods This retrospective study included 128 patients with NF and 133 with STS who underwent preoperative 3-T MRI between May 2023 and June 2025. Two blinded musculoskeletal radiologists independently evaluated predefined imaging features, including fascial tail sign, T2 heterogeneity score, enhancement pattern, perilesional edema, and mean/minimum apparent diffusion coefficient (ADC). Histopathology quantified cellular density, collagen content, and myxoid degeneration. Imaging–pathology correlations were assessed using Spearman analysis and multivariable regression. Diagnostic performance of MRI-only, pathology-only, and combined models was evaluated using logistic regression, receiver operating characteristic (ROC), area under the curve (AUC) analysis, calibration, decision curve analysis, and bootstrap validation. Results T2 heterogeneity correlated positively with myxoid degeneration (ρ = 0.52) and negatively with collagen content (ρ = −0.28) ( p < 0.001). Mean ADC correlated inversely with cellular density (ρ = −0.63, p < 0.001). Mean ADC was significantly higher in NF than STS ( p < 0.001). The combined model outperformed MRI alone (AUC, 0.96 vs 0.89) and demonstrated greater net clinical benefit. Bootstrap validation confirmed stable performance (mean AUC, 0.960; 95% confidence interval, 0.938–0.978). Conclusion Imaging features of NF were associated with underlying stromal and cellular substrates. Integrating imaging features with quantitative pathology biomarkers may support radiology–pathology concordance assessment in selected diagnostically equivocal cases. Key Points Question How can we improve radiology–pathology concordance and reduce interpretative variability for NF cases presenting with suspicious, diagnostically equivocal magnetic resonance imaging findings? Findings T2 heterogeneity and diffusion measurements are associated with stromal composition and cellularity, while predefined MRI features demonstrated substantial-to-excellent inter-reader agreement. Critical relevance statement Embedding explicit AHA criteria in the prompt let an 8B model approach 671B-model agreement (38.91–81.52%), supporting standardized text-based workflows without fine-tuning.