Deep learning-based reconstruction enhances breast DWI image quality and reduces scan time by 29%

Deep learning (DL)-based reconstruction of multi-shot breast DWI significantly improved SNR, CNR, lesion conspicuity, and reduced acquisition time by 29% vs. standard rs-EPI, with gains confirmed in a separate validation cohort, supporting clinical integration.

• DL reconstruction achieved a 29% reduction in acquisition time compared with standard rs-EPI.
• Quantitative improvements in SNR, CNR, and lesion contrast were significant at both b=800 and synthetic b=1500 s/mm² (all p<0.05).
• Qualitative assessments by two radiologists showed higher overall image quality and lesion conspicuity (weighted κ=0.753–0.981); in the validation cohort, ADC values did not differ significantly while image quality gains persisted.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.