Fractal analysis of tumor subregions differentiates benign from malignant breast lesions.

PURPOSE: This study aimed to evaluate the effectiveness of fractal analysis in distinguishing between benign and malignant breast lesions based on tumor subregion complexity and spatial heterogeneity.

MATERIAL AND

METHODS

Breast diffusion-weighted imaging and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data from biopsy- and surgery-confirmed malignant (n = 75) and benign tumors (n = 32) were used to evaluate the whole-lesion heterogeneity and subregional fractal dimension (FD) using habitat imaging. Inter-reader variability in 3D region of interest definition was assessed using the Dice similarity coefficient. Receiver operating characteristic curve (ROC) analysis was conducted to assess the diagnostic performance of studied parameters in differentiating benign from malignant lesions.

RESULTS

The overall Dice similarity coefficient across all lesion classes was 0.86 (95% confidence interval [CI]: 0.79-0.96). The inter-reader reliability was excellent, with intraclass correlation coefficients > 0.90 for FD. Significant differences were observed between malignant and benign lesions in hypervascular cellular FD (p < 0.001) and hypovascular cellular FD (p = 0.011), but not in nonviable tissue FD (p = 0.568). Patients with malignant lesions exhibited significantly higher median heterogeneity of the whole lesion compared to those with benign lesions (p = 0.015). Among the FD features generated by habitat imaging, hypervascular cellular FD had the highest area under the ROC curve (0.925, 95% CI: 0.854-0.969), surpassing the total heterogeneity (0.671, 95% CI: 0.568-0.763). Sensitivity and specificity of 89.3% and 81.8%, respectively, were achieved in distinguishing benign from malignant lesions.

CONCLUSIONS

Subregional FD of tumor habitats demonstrated superior diagnostic accuracy in differentiating between benign malignant breast tumors compared to the heterogeneity of the whole lesion.