Role of Bi-exponential diffusion-weighted model (IVIM) in breast cancer differentiation and comparison of curve fitting methods

Abstract

Objective: The current meta-analysis aimed to analyze all published results about the diagnostic performance of intravoxel incoherent motion (IVIM) to assess the differential diagnosis between malignant and benign breast cancer and comparison of curve fitting methods. Methods: PubMed and Embase databases were systematically searched to find relevant articles published until November2021 in English journals. Studies were evaluated in accordance with inclusion and exclusion standards. We evaluated study heterogeneity and publication bias. Using a bivariate model, aggregated summary values of sensitivity, specificity, and area under the curve were derived for each related parameters of the intravoxel incoherent motion (true diffusivity [D], pseudo-diffusivity [D*], perfusion fraction [f]). Subgroup analysis was employed to investigate the impact of various field strengths and b values. Results: From 25 eligible studies, 2391 lesions (1652 malignant and 653 benign) were included. Publication bias was evident for studies that evaluated parameters of the intravoxel incoherent motion (D, D*, and f). Significant heterogeneity (P 0.05) was present for all parameters. The pooled sensitivity, specificity and area under the curve for f was 0.78, 0.69, and 0.83, respectively. D* was 0.69, 0.63, and 0.72, respectively.