LKCell: Efficient Cell Nuclei Instance Segmentation with Large Convolution Kernels
Ziwei Cui, Jingfeng Yao, Lunbin Zeng, Juan Yang, Wenyu Liu, Xinggang Wang
2024-07-26
Summary
This paper introduces LKCell, a new method for accurately identifying and segmenting cell nuclei in tissue images. It uses large convolution kernels to improve the quality of cell segmentation while being efficient in terms of computation.
What's the problem?
Segmenting cell nuclei in images is important for medical analysis, but traditional methods struggle to balance the need for a large receptive field (to capture more context) with the computational power required. This can lead to lower accuracy in identifying cells and can be resource-intensive, making it difficult to use these methods effectively in practice.
What's the solution?
LKCell solves this problem by using large convolution kernels, which are larger filters that can capture more information from the images without needing as much computational power as previous methods. The researchers adapted pre-trained models that used these large kernels for other tasks and applied them to medical imaging for the first time. They also created a new segmentation decoder that reduces unnecessary complexity while improving performance. As a result, LKCell achieves high accuracy with significantly less computational effort compared to earlier methods.
Why it matters?
This research is important because it enhances the ability to analyze tissue images more effectively, which is crucial for diagnosing diseases and understanding biological processes. By improving cell segmentation techniques, LKCell can help medical professionals make better decisions based on more accurate data, ultimately benefiting patient care.
Abstract
The segmentation of cell nuclei in tissue images stained with the blood dye hematoxylin and eosin (H&E) is essential for various clinical applications and analyses. Due to the complex characteristics of cellular morphology, a large receptive field is considered crucial for generating high-quality segmentation. However, previous methods face challenges in achieving a balance between the receptive field and computational burden. To address this issue, we propose LKCell, a high-accuracy and efficient cell segmentation method. Its core insight lies in unleashing the potential of large convolution kernels to achieve computationally efficient large receptive fields. Specifically, (1) We transfer pre-trained large convolution kernel models to the medical domain for the first time, demonstrating their effectiveness in cell segmentation. (2) We analyze the redundancy of previous methods and design a new segmentation decoder based on large convolution kernels. It achieves higher performance while significantly reducing the number of parameters. We evaluate our method on the most challenging benchmark and achieve state-of-the-art results (0.5080 mPQ) in cell nuclei instance segmentation with only 21.6% FLOPs compared with the previous leading method. Our source code and models are available at https://github.com/hustvl/LKCell.