Related Experiment Video
Updated: Dec 15, 2025

Simplified, High-throughput Analysis of Single-cell Contractility using Micropatterned Elastomers
Published on: April 8, 2022
Image based cellular contractile force evaluation with small-world network inspired CNN: SW-UNet.
Honghan Li1, Daiki Matsunaga1, Tsubasa S Matsui1
1Division of Bioengineering, Graduate School of Engineering Science, Osaka University, Japan, 1-3 Machikaneyama Toyonaka, Osaka, 5608531, Japan.
We developed a machine learning method using SW-UNet to measure cellular contractile force from cell-induced substrate wrinkles. This technique accurately quantifies forces, revealing KRAS-mutated cells exert greater force than wild-type cells.
Area of Science:
- Cellular Biology
- Biophysics
- Machine Learning
Background:
- Cellular contractile force is crucial for biological processes.
- Quantifying cellular forces typically requires complex experimental setups.
- Developing efficient and accurate methods for force measurement is essential.
Purpose of the Study:
- To propose an image-based method for evaluating cellular contractile force using machine learning.
- To develop a novel convolutional neural network (CNN) architecture for wrinkle segmentation.
- To demonstrate the application of this method in comparing cellular forces between different cell types.
Main Methods:
- Utilized a specialized substrate that wrinkles upon cellular contraction to visualize force.
- Developed a new CNN architecture, SW-UNet (small-world U-Net), for accurate wrinkle extraction from microscope images.
- Compared the performance of SW-UNet against 2D-FFT and U-Net for wrinkle segmentation.
Main Results:
- SW-UNet demonstrated superior performance in wrinkle segmentation, with a 4.9x smaller error than 2D-FFT and 2.9x smaller error than U-Net.
- The method successfully visualized and quantified cellular contractile force magnitude and direction.
- U2OS cells with KRAS oncogene mutations exhibited significantly larger contractile forces compared to wild-type cells.
Conclusions:
- The proposed machine learning-based algorithm offers an efficient, automated, and accurate approach for evaluating cellular contractile force.
- SW-UNet provides a significant improvement in wrinkle segmentation accuracy for force evaluation.
- This method facilitates the study of cellular mechanics and the role of contractile forces in diseases like cancer.
Related Concept Videos
Cell-matrix's Response to Mechanical Forces
Anchoring junctions mechanically attach a cell to the...
The Role of Actin and Myosin in Non-muscle Cells

