E1010. Can a B0 Diffusion Image Replace STIR in Breast MRI? Sequence Comparison and Interreader Variability
  1. Jason Gatewood; University of Kansas
  2. Jessica Peterson; University of Kansas
  3. Carissa Walter; University of Kansas
  4. Allison Aripoli; University of Kansas
  5. Ashley Huppe; University of Kansas
  6. Onalisa Winblad; University of Kansas
  7. Katja Pinker-Domenig; Memorial Sloan Kettering Cancer Center
The primary objective of the study was to determine whether diffusion weighted imaging (DWI) low b value images are equivalent to short tau inversion recovery (STIR) images in detecting lesion conspicuity and determining lesion signal characteristics in relation to fibroglandular tissue on breast MRI. The secondary objective was to determine interreader agreement.

Materials and Methods:
72 female patients (mean age ± SD, 48.3 ± 12.5) prospectively underwent standard of care 3T or 1.5T breast MRI with the addition of an investigational DWI sequence, readout segmented echo planar imaging (rs-EPI) combined with simultaneous multislice (SMS). Voxel size and scan time of the DWI and STIR sequences were respectively: 1.2 x 1.2 x 3.0 cm and 4:48 compared with 0.8 x 0.8 x 3.0 cm and 4:45. Lesions were scored by five radiologists (four subspecialty breast radiologist and one breast imaging fellow) independently by completing three questions: Is the conspicuity of the lesion on DWI significantly inferior to STIR? Does the signal characteristics of the lesion on DWI follow the signal characteristics on STIR? Rate DWI artifacts on a scale of 1-4: No artifacts, mild artifacts not affecting assessment of lesion, moderate artifact affecting assessment (3), and nondiagnostic image due to artifact. Statistical analysis included two-sample t-test for lesion comparison and Nelson’s model kappa for interreader agreement testing.

A total of 121 lesions were measured ranging from 0.2 cm - 4.5 cm with a median size of 0.5 cm. From a total of 605 assessments, 537 (88.8%) answered that conspicuity of DWI was not significantly inferior to STIR, 586 (96.9%) agreed that DWI followed signal characteristics on STIR and 595 (98.3%) of 605 exams had no or mild artifacts which did not affect lesion evaluation. There was substantial agreement for lesion conspicuity (89.9%, 0.790, 95% CI 0.76 - 0.82). There was substantial agreement for signal characteristics (95.7%, kappa 0.733, CI 0.69 - 0.77). There was moderate agreement in determining whether there was mild or no artifact versus moderate or greater artifact (97.2%, kappa 0.515, CI 0.29 - 0.75).

DWI has been shown to increase specificity and reduce the number of biopsies in breast MRI. However, this sequence is often omitted from exams to reduce overall scan times. This novel scanning approach presents the opportunity to harness the value of DWI without adding time by replacing STIR with rs-EPI SMS DWI. Our results demonstrate b0 images are noninferior for assessment of lesion conspicuity and lesion signal characteristics as compared with STIR. Our results show rs-EPI SMS DWI is a reliable and reproducible sequence with minimal artifact. With substantial interreader agreement, this sequence may replace STIR imaging in both standard breast MRI and abbreviated breast MRI. This can improve cost effectiveness by reducing false positives and improve patient experience by reducing overall scan time.