Memory Efficient Segmentation of Large Microscopy Images Using Graph-based Neural Networks

Published

June 18, 2024

Author(s)

Atishay Jain, David Laidlaw, Ritambhara Singh, Peter Bajcsy

Abstract

We present a graph neural network (GNN) based framework applied to a large-scale image segmentation task. While deep learning models, like convolutional neural networks (CNNs), have become common for automating image segmentation tasks, they are limited by the image size that can fit to memory of computational hardware. In a GNN framework, large-scale images are converted into graphs allowing us to preserve and represent their global spatial structure. By converting images with hundreds of millions of pixels to graphs with thousands of nodes, we can segment large images using memory-limited computational resources. We compare the performance of GNN- and CNN-based segmentation in terms of accuracy, training time, and required GPU memory. Based on our experiments with microscopy images of biological cells, the GPU-based segmentation used two to three orders-of-magnitude fewer computational resources with only a -3% to +1% change in accuracy. Furthermore, errors due to superpixel generation can be reduced by either using better superpixel generation algorithms or increasing the number of superpixels, thus allowing for improvement in the GNN framework's accuracy. This tradeoff between accuracy and computational cost over CNN models makes the GNN framework attractive for many large-scale image segmentation tasks in biology.

Citation

Microscopy

Issue

3

Pub Type

Journals

Download Paper

https://doi.org/10.1093/jmicro/dfad049

Local Download

Keywords

Graph Neural Networks, Semantic Segmentation, Deep Learning, Computer Vision

Citation

Jain, A. , Laidlaw, D. , Singh, R. and Bajcsy, P. (2024), Memory Efficient Segmentation of Large Microscopy Images Using Graph-based Neural Networks, Microscopy, [online], https://doi.org/10.1093/jmicro/dfad049, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=936141 (Accessed March 12, 2026)