Intravoxel Incoherent Motion and Quantitative Non-Gaussian Diffusion MR Imaging: Evaluation of the Diagnostic and Prognostic Value of Several Markers of Malignant and Benign Breast Lesions

Abstract

Purpose To investigate the performance of integrated approaches that combined intravoxel incoherent motion (IVIM) and non-Gaussian diffusion parameters compared with the Breast Imaging and Reporting Data System (BI-RADS) to establish multiparameter thresholds scores or probabilities by using Bayesian analysis to distinguish malignant from benign breast lesions and their correlation with molecular prognostic factors. Materials and Methods Between May 2013 and March 2015, 411 patients were prospectively enrolled and 199 patients (allocated to training [n = 99] and validation [n = 100] sets) were included in this study. IVIM parameters (flowing blood volume fraction [fIVIM] and pseudodiffusion coefficient [D*]) and non-Gaussian diffusion parameters (theoretical apparent diffusion coefficient [ADC] at b value of 0 sec/mm [ADC] and kurtosis [K]) by using IVIM and kurtosis models were estimated from diffusion-weighted image series (16 b values up to 2500 sec/mm), as well as a synthetic ADC (sADC) calculated by using b values of 200 and 1500 (sADC) and a standard ADC calculated by using b values of 0 and 800 sec/mm (ADC). The performance of two diagnostic approaches (combined parameter thresholds and Bayesian analysis) combining IVIM and diffusion parameters was evaluated and compared with BI-RADS performance. The Mann-Whitney U test and a nonparametric multiple comparison test were used to compare their performance to determine benignity or malignancy and as molecular prognostic biomarkers and subtypes of breast cancer. Results Significant differences were found between malignant and benign breast lesions for IVIM and non-Gaussian diffusion parameters (ADC, K, fIVIM, fIVIM · D*, sADC and ADC; P < .05). Sensitivity and specificity for the validation set by radiologists A and B were as follows: sensitivity, 94.7% and 89.5%, and specificity, 75.0% and 79.2% for sADC, respectively; sensitivity, 94.7% and 96.1%, and specificity, 75.0% and 66.7%, for the combined thresholds approach, respectively; sensitivity, 92.1% and 92.1%, and specificity, 83.3% and 66.7%, for Bayesian analysis, respectively; and sensitivity and specificity, 100% and 79.2%, for BI-RADS, respectively. The significant difference in values of sADC in progesterone receptor status (P = .002) was noted. sADC was significantly different between histologic subtypes (P = .006). Conclusion Approaches that combined various IVIM and non-Gaussian diffusion MR imaging parameters may provide BI-RADS-equivalent scores almost comparable to BI-RADS categories without the use of contrast agents. Non-Gaussian diffusion parameters also differed by biologic prognostic factors. RSNA, 2017 Online supplemental material is available for this article.

Intravoxel incoherent motion diffusion‐weighted MRI of invasive breast cancer: Correlation with prognostic factors and kinetic features acquired with computer‐aided diagnosis

“…Although our measurements were similar to those of most studies investigating IVIM parameters, some recent studies have provided histogram analysis, which could show less fitting artifacts from the parametric map. Second, the range of f value (9.82–48.56%) was relatively high as compared to those (0–46.74%) from the literature . Those high values likely reflect residual tissue diffusion, which can be attributed to the heterogeneity of the tumors.…”

“…Second, the range of f value (9.82-48.56%) was relatively high as compared to those (0-46.74%) from the literature. [5][6][7][8][9][10][11]31 Those high values likely reflect residual tissue diffusion, which can be attributed to the heterogeneity of the tumors. However, according to a recent study using histogram analysis, 6 the maximum f value of invasive breast cancers was reported as 46.74%, which was similar to our study result and we used the median value rather than the mean value in our statistical analysis to avoid average trap.…”

“…Through IVIM, not only can pure diffusion (D) be separated, but perfusion parameters such as pseudodiffusion (Dp) and the microvascular volume fraction (f) can also be acquired without the use of a contrast agent . When investigating the application of IVIM modeling in breast lesions, most studies have focused on the additional role of IVIM parameters to differentiate between benign and malignant breast lesions . Recently, some studies have investigated the association between the IVIM parameters and the prognostic factors of breast cancers .…”

“…4 When investigating the application of IVIM modeling in breast lesions, most studies have focused on the additional role of IVIM parameters to differentiate between benign and malignant breast lesions. 2,[5][6][7][8][9] Recently, some studies have investigated the association between the IVIM parameters and the prognostic factors of breast cancers. 6,10,11 However, there has been little agreement between the results.…”

A Meta-analysis of the Diagnostic Performance of Diffusion MRI for Breast Lesion Characterization

“…Results of quality assessment for risk of bias and concerns regarding applicability of included studies. Quality assessment of diagnostic accuracy studies (QUADAS-2) scores for each category are expressed by percentages of studies that have a low, high, or unclear risk of bias.IVIM parameters and Iima et al(91) reported a combination of f, ADC, and diffusion kurtosis coefficient K.Jiang et al (27) also reported a combination of fractional anisotropy and the maximal anisotropy index.…”