TY - GEN
T1 - BoxShrink
T2 - 1st International Workshop on Medical Image Learning with Limited and Noisy Data, MILLanD 2022, held in conjunction with 25th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2022
AU - Gröger, Michael
AU - Borisov, Vadim
AU - Kasneci, Gjergji
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - One of the core challenges facing the medical image computing community is fast and efficient data sample labeling. Obtaining fine-grained labels for segmentation is particularly demanding since it is expensive, time-consuming, and requires sophisticated tools. On the contrary, applying bounding boxes is fast and takes significantly less time than fine-grained labeling, but does not produce detailed results. In response, we propose a novel framework for weakly-supervised tasks with the rapid and robust transformation of bounding boxes into segmentation masks without training any machine learning model, coined BoxShrink. The proposed framework comes in two variants – rapid-BoxShrink for fast label transformations, and robust-BoxShrink for more precise label transformations. An average of four percent improvement in IoU is found across several models when being trained using BoxShrink in a weakly-supervised setting, compared to using only bounding box annotations as inputs on a colonoscopy image data set. We open-sourced the code for the proposed framework and published it online.
AB - One of the core challenges facing the medical image computing community is fast and efficient data sample labeling. Obtaining fine-grained labels for segmentation is particularly demanding since it is expensive, time-consuming, and requires sophisticated tools. On the contrary, applying bounding boxes is fast and takes significantly less time than fine-grained labeling, but does not produce detailed results. In response, we propose a novel framework for weakly-supervised tasks with the rapid and robust transformation of bounding boxes into segmentation masks without training any machine learning model, coined BoxShrink. The proposed framework comes in two variants – rapid-BoxShrink for fast label transformations, and robust-BoxShrink for more precise label transformations. An average of four percent improvement in IoU is found across several models when being trained using BoxShrink in a weakly-supervised setting, compared to using only bounding box annotations as inputs on a colonoscopy image data set. We open-sourced the code for the proposed framework and published it online.
KW - Colonoscopy
KW - Deep neural networks
KW - Segmentation
KW - Weakly-supervised learning
UR - http://www.scopus.com/inward/record.url?scp=85140474936&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-16760-7_7
DO - 10.1007/978-3-031-16760-7_7
M3 - Conference contribution
AN - SCOPUS:85140474936
SN - 9783031167591
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 65
EP - 75
BT - Medical Image Learning with Limited and Noisy Data - 1st International Workshop, MILLanD 2022, Held in Conjunction with MICCAI 2022, Proceedings
A2 - Zamzmi, Ghada
A2 - Antani, Sameer
A2 - Rajaraman, Sivaramakrishnan
A2 - Xue, Zhiyun
A2 - Bagci, Ulas
A2 - Linguraru, Marius George
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 22 September 2022 through 22 September 2022
ER -