TY - GEN
T1 - Pre-procedural Planning Algorithm for Mammogram-guided Wire-Localization of Breast Microcalcifications Lesions
AU - Suppuang, Chavisa
AU - Wiratkapun, Cholatip
AU - Suthakorn, Jackrit
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Wire localization under conventional mammography presents challenges in translating two-dimensional imaging into precise spatial targeting due to compression variability between craniocaudal (CC) and mediolateral oblique (MLO) views. This work presents a computer-assisted planning algorithm combining domain-adapted UNet segmentation with clinician-decision mimicking algorithmic reasoning for optimal wire placement recommendations. The system employs a UNet segmentation model for microcalcifications adapted to institutional data and mammographic image segmentation pipeline for breast feature segmentation using gradient-based methods. The algorithm calculates minimum lesion-to-surface distances in CC and MLO views, recommending optimal compression approaches based on shortest access paths, with simplified 3D visualization through geometric decompression and view correlation. The novelty of this approach lies in providing an automated, pre-procedural planning solution that replicates radiologists' spatial reasoning. Retrospective evaluation on 30 cases demonstrates 83.3% agreement with expert clinical decisions, 86.7% successful completion rate, and 94 seconds mean processing time on general-purpose workstation. The system provides clinically acceptable decision support while preserving physician autonomy in mammographic wire localization procedures.
AB - Wire localization under conventional mammography presents challenges in translating two-dimensional imaging into precise spatial targeting due to compression variability between craniocaudal (CC) and mediolateral oblique (MLO) views. This work presents a computer-assisted planning algorithm combining domain-adapted UNet segmentation with clinician-decision mimicking algorithmic reasoning for optimal wire placement recommendations. The system employs a UNet segmentation model for microcalcifications adapted to institutional data and mammographic image segmentation pipeline for breast feature segmentation using gradient-based methods. The algorithm calculates minimum lesion-to-surface distances in CC and MLO views, recommending optimal compression approaches based on shortest access paths, with simplified 3D visualization through geometric decompression and view correlation. The novelty of this approach lies in providing an automated, pre-procedural planning solution that replicates radiologists' spatial reasoning. Retrospective evaluation on 30 cases demonstrates 83.3% agreement with expert clinical decisions, 86.7% successful completion rate, and 94 seconds mean processing time on general-purpose workstation. The system provides clinically acceptable decision support while preserving physician autonomy in mammographic wire localization procedures.
KW - Computer-assisted planning
KW - Image processing
KW - Mammography
KW - Wire localization
UR - https://www.scopus.com/pages/publications/105037458679
U2 - 10.1109/I-CREATE67590.2025.11478156
DO - 10.1109/I-CREATE67590.2025.11478156
M3 - Conference contribution
AN - SCOPUS:105037458679
T3 - 2025 International Convention on Rehabilitation Engineering and Assistive Technology, i-CREATe 2025 - Conference Proceedings
BT - 2025 International Convention on Rehabilitation Engineering and Assistive Technology, i-CREATe 2025 - Conference Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 18th International Convention on Rehabilitation Engineering and Assistive Technology, i-CREATe 2025
Y2 - 24 November 2025 through 26 November 2025
ER -