Towards Accurate and Efficient Pan-cancer Screening

Linshan Wu, Jia-Xin Zhuang

Keywords: Pan-cancer screening

TL;DR: An accurate and efficient framework for pan-cancer screening

Abstract: Pan-cancer screening is challenging due to the difficulty of localizing tiny tumors within large CT volumes. The efficiency also plays an important role in large-scale routine physical examination. To this end, we propose an accurate and efficient framework for pan-cancer screening. Our model employs a classification model to localize the diseased regions and a segmentation model to precisely segment the lesions. Specifically, we first crop a group of sub-volumes from the whole CT volume and select those with lesions for segmentation. We lead the MICCAI FLARE25 pan-cancer challenge, with 59\% DSC and 52\% NSD on the validation leaderboard.

Submission Number: 16