Multi-Modality Cell Segmentation based on nnU-Net Pipeline

Haotian Lu, Jinghao Feng, Zelin Peng

Keywords: Medical Image Segmentation, Multi-Modality, Semi-supervised Learning

TL;DR: A technical report of NeurIPS CellSeg 2022

Abstract: Cell segmentation is an important upstream task in Medical Image Analysis. Recent years, data-driven deep learning methods made ground-breaking achievements. In this challenge, multi-modal and partially-labeled dataset is provided. On the dataset, this paper proposes a multi-modality cell segmentation framework based on nnU-Net pipeline and iterative self-training method. Our model reaches 0.6101 mean F1 score on tuning set.