TY - GEN
T1 - DisAsymNet
T2 - 26th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2023
AU - Wang, Xin
AU - Tan, Tao
AU - Gao, Yuan
AU - Han, Luyi
AU - Zhang, Tianyu
AU - Lu, Chunyao
AU - Beets-Tan, Regina
AU - Su, Ruisheng
AU - Mann, Ritse
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.
PY - 2023
Y1 - 2023
N2 - Asymmetry is a crucial characteristic of bilateral mammograms (Bi-MG) when abnormalities are developing. It is widely utilized by radiologists for diagnosis. The question of “what the symmetrical Bi-MG would look like when the asymmetrical abnormalities have been removed ?” has not yet received strong attention in the development of algorithms on mammograms. Addressing this question could provide valuable insights into mammographic anatomy and aid in diagnostic interpretation. Hence, we propose a novel framework, DisAsymNet, which utilizes asymmetrical abnormality transformer guided self-adversarial learning for disentangling abnormalities and symmetric Bi-MG. At the same time, our proposed method is partially guided by randomly synthesized abnormalities. We conduct experiments on three public and one in-house dataset, and demonstrate that our method outperforms existing methods in abnormality classification, segmentation, and localization tasks. Additionally, reconstructed normal mammograms can provide insights toward better interpretable visual cues for clinical diagnosis. The code will be accessible to the public.
AB - Asymmetry is a crucial characteristic of bilateral mammograms (Bi-MG) when abnormalities are developing. It is widely utilized by radiologists for diagnosis. The question of “what the symmetrical Bi-MG would look like when the asymmetrical abnormalities have been removed ?” has not yet received strong attention in the development of algorithms on mammograms. Addressing this question could provide valuable insights into mammographic anatomy and aid in diagnostic interpretation. Hence, we propose a novel framework, DisAsymNet, which utilizes asymmetrical abnormality transformer guided self-adversarial learning for disentangling abnormalities and symmetric Bi-MG. At the same time, our proposed method is partially guided by randomly synthesized abnormalities. We conduct experiments on three public and one in-house dataset, and demonstrate that our method outperforms existing methods in abnormality classification, segmentation, and localization tasks. Additionally, reconstructed normal mammograms can provide insights toward better interpretable visual cues for clinical diagnosis. The code will be accessible to the public.
KW - Asymmetric transformer
KW - Bilateral mammogram
KW - Disentanglement
KW - Self-adversarial learning
KW - Synthesis
UR - http://www.scopus.com/inward/record.url?scp=85174727025&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-43990-2_6
DO - 10.1007/978-3-031-43990-2_6
M3 - Conference contribution
AN - SCOPUS:85174727025
SN - 9783031439896
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 57
EP - 67
BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings
A2 - Greenspan, Hayit
A2 - Greenspan, Hayit
A2 - Madabhushi, Anant
A2 - Mousavi, Parvin
A2 - Salcudean, Septimiu
A2 - Duncan, James
A2 - Syeda-Mahmood, Tanveer
A2 - Taylor, Russell
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 8 October 2023 through 12 October 2023
ER -