Lung Nodule Segmentation Network with Self-Supervised Learning and Attention Mechanisms

Eunah Jung, Seungmin Chou, Changwon Lim

Keywords: Pulmonary Nodule Detection and Segmentation, 3D Segmentation

Abstract: Pulmonary nodule detection is one of the most important tasks for early lung cancer diagnosis. Especially, end-to-end methods for multi-tasking, including pulmonary nodule detection, false positive detection, and segmentation have been widely used based on supervised learning, leading to significant improvement in performance when detecting pulmonary nodules. However, those methods with confined environments were not able to exploit the representative features comprehensively. Therefore, some self-supervised methods have been proposed to handle the raw dataset. However, they were merely applied to each task, missing rich features of the end-to-end framework. In this paper, we propose a novel adaptation of self-supervised learning to a multi-tasking framework. Additionally, we employed other attention methods, such as Convolutional Block Attention Module(CBAM), and Quartet Attention Mechanism(QAM) to further enhance the performance without significantly in- creasing the number of parameters to learn.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 8618