Image Segmentation with Topological Priors

TL;DR

This work uses topological priors before and during the deep neural network training procedure to solve segmentation tasks with topological priors and found that incorporating topological information into the classical U - Net model performed significantly better.

Abstract

Solving segmentation tasks with topological priors proved to make fewer errors in fine-scale structures. In this work, we use topological priors both before and during the deep neural network training procedure. We compared the results of the two approaches with simple segmentation on various accuracy metrics and the Betti number error, which is directly related to topological correctness, and discovered that incorporating topological information into the classical UNet model performed significantly better. We conducted experiments on the ISBI EM segmentation dataset.

Figures (7)

• Figure 1: Topological input image processing for segmentation
• Figure 2: Image and corresponding Mask
• Figure 3: Architecture of U-Net
• Figure 4: Persistence diagrams for 5 (left) and 95 (right) epochs
• Figure 5: Predictions without topological priors (middle row) and with topoloss (bottom row)