Three-compartment spectral diffusion analysis for breast cancer magnetic resonance imaging

To examine the diagnostic performance of a three-compartment diffusion model with the fixed cut-off diffusion coefficient (D) using magnetic resonance spectral diffusion analysis for differentiating between invasive ductal carcinoma (IDC) and ductal carcinoma in situ (DCIS) and compare the conventional apparent D (ADC), and mean kurtosis (MK), with the tissue D (D_IVIM), perfusion D (D*_IVIM), and perfusion fraction (f_IVIM) calculated by conventional intravoxel incoherent motion.

This retrospective study included women who underwent breast MRI with eight b-value diffusion-weighted imaging between February 2019 and March 2022. Spectral diffusion analysis was performed; very-slow, cellular, and perfusion compartments were defined using cut-off Ds of 0.1 × 10^-3 and 3.0 × 10^-3 mm²/s (static water D). The mean D (D_s, D_c, D_p, respectively) and fraction F (F_s, F_c, F_p, respectively) for each compartment were calculated. ADC and MK values were also calculated; receiver operating characteristic analyses were performed.

Histologically confirmed 132 ICD and 62 DCIS (age range 31-87 [53 ± 11] years) were evaluated. The areas under the curve (AUCs) for ADC, MK, D_IVIM, D*_IVIM, f_IVIM, D_s, D_c, D_p, F_s, F_c, and F_p were 0.77, 0.72, 0.77, 0.51, 0.67, 0.54, 0.78, 0.51, 0.57, 0.54, and 0.57, respectively. The AUCs for the model combining very-slow and cellular compartments and the model combining the three compartments were 0.81 each, slightly and significantly higher than for ADC, D_IVIM, and D_c (P = 0.09-0.14); and MK (P < 0.05), respectively.

Three-compartment model analysis using the diffusion spectrum accurately differentiated IDC from DCIS; however, it was not superior to ADC and D_IVIM. The diagnostic performance of MK was lower than that of the three-compartment model.