Abstract
We study the problem of training an accurate deep learning mitosis detection model with only point annotations. To address this challenging label-efficient deep learning problem, we propose a novel contextual prior constraint mechanism and spatial area constrained loss to generate the reference ground truth for segmentation and to restrain incorrectly predicted pixels, respectively. The spatial area constraint mechanism is not strictly cast at the pixel-level and restrains the mitosis and non-mitosis areas as positive/negative bags under the framework of multiple instance learning. The experimental results show that our contextual prior mechanism with PSPNet as the segmentation baseline achieves state-of-the-art performance with an F-score of 69.92%, 56.22%, and 85.29% on the mitosis detection task of AMIDA 2013, ICPR MITOSIS 2014, and point-annotated ICPR MITOSIS 2012, respectively. Especially, using our spatial area constraint mechanism and reference ground truth, the detection result on point-annotated ICPR MITOSIS 2012 even outperforms the result using the same backbone network with pixel-level annotations. The experimental results demonstrate the advancement and effectiveness of our proposed method. In addition, they indicate that our work can definitely improve the performance of mitosis detection on point-annotated datasets and be extended to other medical image analysis tasks with limited annotations.
Highlights
Mitosis count is a crucial factor in breast cancer diagnosis and mitosis detection remains a significant issue in medical image analysis
4) NUMERICAL RESULTS The experimental results show that using the contextual prior constraint method with pyramid scene parsing networks (PSPNet)(50) as the segmentation network and spatial area constrained loss, we achieve stateof-the-art performance of mitosis detection on the AMIDA 2013 test set with an F-score of 69.92%, outperforming the previous best result, i.e., SegMitos [32] with an F-score of 67.28% by 2.64%
3) NUMERICAL RESULTS The experimental results show that using our CPCN with PSPNet(50) as the segmentation network and adding the spatial area constrained term to weighted cross-entropy with the weight value of 17, we achieve state-of-the-art results of mitosis detection on the test set of ICPR MITOSIS 2014 with an F-score of 56.22%
Summary
Mitosis count is a crucial factor in breast cancer diagnosis and mitosis detection remains a significant issue in medical image analysis. Deep learning has achieved excellent performance on computer vision in recent years and mitosis detection can be considered a specific task of object detection or semantic segmentation, which are both priority issues of computer vision. It is reasonable to approach the mitosis detection task via deep learning methods. Supervised semantic segmentation networks need pixel-level annotations and object detection methods require the bounding boxes of the objects as the ground truth (GT). To achieve mitosis detection on these point-annotated datasets using fully supervised methods, we propose the contextual prior constrained
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
