A comparative study of semi- and self-supervised semantic segmentation of biomedical microscopy data

TL;DR

This paper compares semi- and self-supervised CNN training methods for biomedical microscopy image segmentation, aiming to reduce the dependency on large labeled datasets in biomedical image analysis.

Contribution

It introduces and evaluates adapted semi- and self-supervised classification methods for biomedical image segmentation, highlighting their effectiveness with limited labels.

Findings

Semi- and self-supervised methods perform comparably to supervised training with fewer labels.

These methods reduce the need for extensive labeled datasets in biomedical segmentation.

The adapted methods show promise for practical biomedical imaging applications.

Abstract

In recent years, Convolutional Neural Networks (CNNs) have become the state-of-the-art method for biomedical image analysis. However, these networks are usually trained in a supervised manner, requiring large amounts of labelled training data. These labelled data sets are often difficult to acquire in the biomedical domain. In this work, we validate alternative ways to train CNNs with fewer labels for biomedical image segmentation using. We adapt two semi- and self-supervised image classification methods and analyse their performance for semantic segmentation of biomedical microscopy images.