Diffusion tensor imaging for characterizing tumor microstructure and improving diagnostic performance on breast MRI: a prospective observational study

Abstract

Background: Diffusion-weighted imaging (DWI) can increase breast MRI diagnostic specificity due to the tendency of malignancies to restrict diffusion. Diffusion tensor imaging (DTI) provides further information over conventional DWI regarding diffusion directionality and anisotropy. Our study evaluates DTI features of suspicious breast lesions detected on MRI to determine the added diagnostic value of DTI for breast imaging.

Methods: With IRB approval, we prospectively enrolled patients over a 3-year period who had suspicious (BI-RADS category 4 or 5) MRI-detected breast lesions with histopathological results. Patients underwent multiparametric 3 T MRI with dynamic contrast-enhanced (DCE) and DTI sequences. Clinical factors (age, menopausal status, breast density, clinical indication, background parenchymal enhancement) and DCE-MRI lesion parameters (size, type, presence of washout, BI-RADS category) were recorded prospectively by interpreting radiologists. DTI parameters (apparent diffusion coefficient [ADC], fractional anisotropy [FA], axial diffusivity [λ₁], radial diffusivity [(λ₂ + λ₃)/2], and empirical difference [λ₁ - λ₃]) were measured retrospectively. Generalized estimating equations (GEE) and least absolute shrinkage and selection operator (LASSO) methods were used for univariate and multivariate logistic regression, respectively. Diagnostic performance was internally validated using the area under the curve (AUC) with bootstrap adjustment.

Results: The study included 238 suspicious breast lesions (95 malignant, 143 benign) in 194 women. In univariate analysis, lower ADC, axial diffusivity, and radial diffusivity were associated with malignancy (OR = 0.37-0.42 per 1-SD increase, p < 0.001 for each), as was higher FA (OR = 1.45, p = 0.007). In multivariate analysis, LASSO selected only ADC (OR = 0.41) as a predictor for a DTI-only model, while both ADC (OR = 0.41) and FA (OR = 0.88) were selected for a model combining clinical and imaging parameters. Post-hoc analysis revealed varying association of FA with malignancy depending on the lesion type. The combined model (AUC = 0.81) had a significantly better performance than Clinical/DCE-MRI-only (AUC = 0.76, p < 0.001) and DTI-only (AUC = 0.75, p = 0.002) models.

Conclusions: DTI significantly improves diagnostic performance in multivariate modeling. ADC is the most important diffusion parameter for distinguishing benign and malignant breast lesions, while anisotropy measures may help further characterize tumor microstructure and microenvironment.

Keywords: Apparent diffusion coefficient (ADC); Breast MRI; Diagnosis; Diffusion tensor imaging (DTI); Dynamic contrast-enhanced (DCE) MRI; False positives; Fractional anisotropy (FA); Suspicious lesions.

Fig. 1

Malignant mass detected in a 47-year-old patient undergoing MRI to evaluate newly diagnosed cancer. a DCE post-contrast subtraction image demonstrates an additional 20-mm round mass with irregular margins in the posterior right breast 6 o’clock (arrow), assigned a BI-RADS category 4. DTI-derived parametric maps of b apparent diffusion coefficient (ADC), c fractional anisotropy, and eigenvalues d λ₁, e λ₂, and f λ₃ are shown for the lesion regions overlaid in color on the b = 800 s/mm² image. ADC, λ₁, λ₂, and λ₃ are in units of 10⁻³ (mm²/s). The mass demonstrated low ADC (mean ADC = 1.09 × 10⁻³ mm²/s) with FA = 0.18, λ₁ = 1.27 × 10⁻³ mm²/s, λ₂ = 1.11 × 10⁻³ mm²/s, and λ₃ = 0.87 × 10⁻³ mm²/s. Biopsy revealed a malignant grade 1 invasive ductal carcinoma

Fig. 2

Benign mass detected in a 45-year-old patient undergoing MRI for high risk screening. a DCE post-contrast image demonstrates an 18-mm irregular mass with irregular margins in the left breast 7 o’clock (arrow), assigned a BI-RADS category 4. DTI-derived parametric maps of b apparent diffusion coefficient (ADC), c fractional anisotropy, and eigenvalues d λ₁, e λ₂, and f λ₃ are shown for the lesion regions overlaid in color on the b = 800 s/mm² image. ADC, λ₁, λ₂, and λ₃ are in units of 10⁻³ (mm²/s). The mass demonstrated high ADC (mean ADC = 2.00 × 10⁻³ mm²/s) and very low FA (FA = 0.10), with λ₁ = 2.20 × 10⁻³ mm²/s, λ₂ = 2.00 × 10⁻³ mm²/s, and λ₃ = 1.80 × 10⁻³ mm²/s. Ultrasound-guided biopsy revealed benign fibroadenoma

Fig. 3

Malignant non-mass enhancement detected in a 58-year-old patient undergoing MRI for high risk screening. a DCE post-contrast image demonstrates a 33-mm linear heterogeneous non-mass enhancement in the posterior left breast 2 o’clock (arrow), assigned a BI-RADS category 4. DTI-derived parametric maps of b apparent diffusion coefficient (ADC), c fractional anisotropy, and eigenvalues d λ₁, e λ₂, and f λ₃ are shown for the lesion regions overlaid in color on the b = 800 s/mm² image. ADC, λ₁, λ₂, and λ₃ are in units of 10⁻³ (mm²/s). The lesion demonstrated moderate ADC (mean ADC = 1.47 × 10⁻³ mm²/s) with FA = 0.25, λ₁ = 1.83 × 10⁻³ mm²/s, λ₂ = 1.47 × 10⁻³ mm²/s, and λ₃ = 1.11 × 10⁻³ mm²/s. MR-guided biopsy revealed DCIS

Fig. 4

Benign non-mass enhancement detected in a 40-year-old patient undergoing MRI to evaluate newly diagnosed cancer. a DCE post-contrast image demonstrates a 20-mm focal heterogeneous non-mass enhancement in the middle right breast 9 o’clock (arrow), assigned a BI-RADS category 4. DTI-derived parametric maps of b apparent diffusion coefficient (ADC), c fractional anisotropy, and eigenvalues d λ₁, (e) λ₂, and (f) λ₃ are shown for the lesion regions overlaid in color on the b = 800 s/mm² image. ADC, λ₁, λ₂, and λ₃ are in units of 10⁻³ (mm²/s). The lesion demonstrated moderate ADC (mean ADC = 1.56 × 10⁻³ mm²/s) and high FA (FA = 0.39), with λ₁ = 2.13 × 10⁻³ mm²/s, λ₂ = 1.58 × 10⁻³ mm²/s, and λ₃ = 0.97 × 10⁻³ mm²/s. MR-guided biopsy revealed benign usual ductal hyperplasia

Fig. 5

Cross-validated ROC curves for Clinical/DCE-MRI-only, DTI-only, and Clinical/DCE-MRI+DTI models to discriminate malignant and benign lesions. The bootstrap-adjusted AUC estimates were 0.76 (95% CI 0.71–0.83), 0.75 (95% CI 0.68–0.82), and 0.81 (95% CI 0.78–0.88), respectively. The Clinical/DCE-MRI+DTI model had a significantly higher AUC than the Clinical/DCE-MRI model (p < 0.001) and DTI-only model (p = 0.002)

Fig. 6

ROC curves for individual DTI parameters within the lesion subgroups defined by type or size. AUC estimates were significantly higher in masses than non-masses for both ADC (AUC 0.84 [95% CI 0.77–0.91] vs. 0.63 [95% CI 0.52–0.75], p = 0.002) and FA (AUC 0.69 [95% CI 0.59–0.79] vs. 0.50 [95% CI 0.40–0.61], p = 0.013). By contrast, AUC estimates were not significantly different between large (≥ 1 cm) and small (< 1 cm) lesions for both ADC (AUC 0.79 [95% CI 0.71–0.86] vs. 0.69 [95% CI 0.57–0.81], p = 0.18) and FA (AUC 0.64 [95% CI 0.55–0.73] vs. 0.62 [95% CI 0.49–0.75] p = 0.80)

Fig. 7

Comparison between the cross-validated ROC curve for Clinical/DCE-MRI+DTI+interactions model and curves for Clinical/DCE-MRI-only and Clinical/DCE-MRI+DTI models. The interaction terms were size × ADC, type × ADC, size × FA, and type × FA and allowed the model to have different odds ratios for ADC and FA for each size × type subgroup. The bootstrap-adjusted AUC estimates were 0.76 (95% CI 0.71–0.83), 0.81 (95% CI 0.78–0.88), and 0.85 (95% CI 0.82–0.90), respectively. The Clinical/DCE-MRI+DTI+interactions model had a significantly higher AUC than the Clinical/DCE-MRI+DTI model (p = 0.018)