ESTRO 2024 - Abstract Book

S3053

Physics - Autosegmentation

ESTRO 2024

Figure 1: results of our method and baseline method (LayerCAM) for the (i) AutoPET (FDG-PET/CT), (ii) MosMed (COVID 19 thoracic CT) and (iii) Duke breast cancer (MRI) datasets.

Conclusion:

We propose a new method to tackle weakly supervised delineation with volumetric images. Our results demonstrate that our method surpasses benchmark tools, offering promising avenues for enhancing tumor/lesion delineation. This method can lead to fully automatic delineation pipelines that can be used in radiotherapy treatment planning and biomarker discovery.

Keywords: deep learning, tumors and lesions segmentation

References: [1] He, Kaiming, et al. "Identity mappings in deep residual networks." Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part IV 14. Springer International Publishing, 2016. [2] Saha, Ashirbani, et al. "A machine learning approach to radiogenomics of breast cancer: a study of 922 subjects and 529 DCE-MRI features." British journal of cancer 119.4 (2018): 508-516. [3] Sergios Gatidis and Thomas Kuestner. A whole-body FDG-PET/CT dataset with manually annotated tumor lesions, 2022. [4] Sergey P Morozov, AE Andreychenko, NA Pavlov, AV Vladzymyrskyy, NV Ledikhova, VA Gombolevskiy, Ivan A Blokhin, PB Gelezhe, AV Gonchar, and V Yu Chernina. Mosmeddata: Chest ct scans with covid-19 related findings dataset. arXiv preprint arXiv:2005.06465, 2020. [5] Jiang, Peng-Tao, et al. "Layercam: Exploring hierarchical class activation maps for localization." IEEE Transactions on Image Processing 30 (2021): 5875-5888.