MRI-guided vacuum-assisted breast biopsy (VABB) is used for suspicious breast cancer (BC) lesions which are detectable only with MRI: because the high sensitivity but limited specificity of breast MRI it is a fundamental tool in breast imaging divisions. We analyse our experience of MRI-guided VABB and critically discuss the potentialities of diffusion-weighted imaging (DWI) and artificial intelligence (AI) in this matter. We retrospectively analysed a population of consecutive women underwent VABB at our tertiary referral BC centre from 01/2011 to 01/2019. Reference standard was histological diagnosis or at least 1-year negative follow-up. McNemar, Mann-Whitney and χ2 tests at 95% level of significance were used as statistical exams. 217 women (mean age = 52, 18-72 years) underwent MRI-guided VABB; 11 were excluded and 208 MRI-guided VABB lesions were performed: 34/208 invasive carcinomas, 32/208 DCIS, 8/208 LCIS, 3/208 high-risk lesions and 131/208 benign lesions were reported. Accuracy of MRI-guided VABB was 97%. The predictive features for malignancy were mass with irregular shape (OR 8.4; 95% CI 0.59-31.6), size of the lesion (OR 4.4; 95% CI 1.69-9.7) and mass with irregular/spiculated margins (OR 5.4; 95% CI 6.8-31.1). Six-month follow-up showed 4 false-negative cases (1.9%). Invasive BC showed a statistically significant higher hyperintense signal at DWI compared to benign lesions (p = 0.03). No major complications occurred. MR-guided VABB showed high accuracy. Benign-concordant lesions should be followed up with breast MRI in 6-12 months due to the risk of false-negative results. DWI and AI applications showed potential benefit as support tools for radiologists.
Keywords: Artificial intelligence; Breast MRI; Breast cancer; Breast imaging; Diffusion-weighted imaging; Oncology; Radiology; Vacuum-assisted breast biopsy.