TY - JOUR
T1 - Factors Affecting Image Quality and Lesion Evaluability in Breast Diffusion-weighted MRI
T2 - Observations from the ECOG-ACRIN Cancer Research Group Multisite Trial (A6702)
AU - Whisenant, Jennifer G.
AU - Romanoff, Justin
AU - Rahbar, Habib
AU - Kitsch, Averi E.
AU - Harvey, Sara M.
AU - Moy, Linda
AU - Demartini, Wendy B.
AU - Dogan, Basak E.
AU - Yang, Wei T.
AU - Wang, Lilian C.
AU - Joe, Bonnie N.
AU - Wilmes, Lisa J.
AU - Hylton, Nola M.
AU - Oh, Karen Y.
AU - Tudorica, Luminita A.
AU - Neal, Colleen H.
AU - Malyarenko, Dariya I.
AU - McDonald, Elizabeth S.
AU - Comstock, Christopher E.
AU - Yankeelov, Thomas E.
AU - Chenevert, Thomas L.
AU - Partridge, Savannah C.
N1 - Publisher Copyright:
© 2020 Society of Breast Imaging 2020. All rights reserved.
PY - 2021/1/1
Y1 - 2021/1/1
N2 - Objective: The A6702 multisite trial confirmed that apparent diffusion coefficient (ADC) measures can improve breast MRI accuracy and reduce unnecessary biopsies, but also found that technical issues rendered many lesions non-evaluable on diffusion-weighted imaging (DWI). This secondary analysis investigated factors affecting lesion evaluability and impact on diagnostic performance. Methods: The A6702 protocol was IRB-approved at 10 institutions; participants provided informed consent. In total, 103 women with 142 MRI-detected breast lesions (BI-RADS assessment category 3, 4, or 5) completed the study. DWI was acquired at 1.5T and 3T using a four b-value, echo-planar imaging sequence. Scans were reviewed for multiple quality factors (artifacts, signal-to-noise, misregistration, and fat suppression); lesions were considered non-evaluable if there was low confidence in ADC measurement. Associations of lesion evaluability with imaging and lesion characteristics were determined. Areas under the receiver operating characteristic curves (AUCs) were compared using bootstrapping. Results: Thirty percent (42/142) of lesions were non-evaluable on DWI; 23% (32/142) with image quality issues, 7% (10/142) with conspicuity and/or localization issues. Misregistration was the only factor associated with non-evaluability (P = 0.001). Smaller (≤10 mm) lesions were more commonly non-evaluable than larger lesions (p <0.03), though not significant after multiplicity correction. The AUC for differentiating benign and malignant lesions increased after excluding non-evaluable lesions, from 0.61 (95% CI: 0.50-0.71) to 0.75 (95% CI: 0.65-0.84). Conclusion: Image quality remains a technical challenge in breast DWI, particularly for smaller lesions. Protocol optimization and advanced acquisition and post-processing techniques would help to improve clinical utility.
AB - Objective: The A6702 multisite trial confirmed that apparent diffusion coefficient (ADC) measures can improve breast MRI accuracy and reduce unnecessary biopsies, but also found that technical issues rendered many lesions non-evaluable on diffusion-weighted imaging (DWI). This secondary analysis investigated factors affecting lesion evaluability and impact on diagnostic performance. Methods: The A6702 protocol was IRB-approved at 10 institutions; participants provided informed consent. In total, 103 women with 142 MRI-detected breast lesions (BI-RADS assessment category 3, 4, or 5) completed the study. DWI was acquired at 1.5T and 3T using a four b-value, echo-planar imaging sequence. Scans were reviewed for multiple quality factors (artifacts, signal-to-noise, misregistration, and fat suppression); lesions were considered non-evaluable if there was low confidence in ADC measurement. Associations of lesion evaluability with imaging and lesion characteristics were determined. Areas under the receiver operating characteristic curves (AUCs) were compared using bootstrapping. Results: Thirty percent (42/142) of lesions were non-evaluable on DWI; 23% (32/142) with image quality issues, 7% (10/142) with conspicuity and/or localization issues. Misregistration was the only factor associated with non-evaluability (P = 0.001). Smaller (≤10 mm) lesions were more commonly non-evaluable than larger lesions (p <0.03), though not significant after multiplicity correction. The AUC for differentiating benign and malignant lesions increased after excluding non-evaluable lesions, from 0.61 (95% CI: 0.50-0.71) to 0.75 (95% CI: 0.65-0.84). Conclusion: Image quality remains a technical challenge in breast DWI, particularly for smaller lesions. Protocol optimization and advanced acquisition and post-processing techniques would help to improve clinical utility.
KW - apparent diffusion coefficient (ADC)
KW - artifacts
KW - breast magnetic resonance imaging (MRI)
KW - diagnostic performance
KW - multicenter trial
UR - http://www.scopus.com/inward/record.url?scp=85100790106&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85100790106&partnerID=8YFLogxK
U2 - 10.1093/jbi/wbaa103
DO - 10.1093/jbi/wbaa103
M3 - Article
AN - SCOPUS:85100790106
SN - 2631-6110
VL - 3
SP - 44
EP - 56
JO - Journal of Breast Imaging
JF - Journal of Breast Imaging
IS - 1
ER -