Cell Detection Using Extremal Regions in a Semisupervised Learning Framework.

Nisha Ramesh,Tolga Tasdizen,Ting Liu

doi:10.1155/2017/4080874

Nisha Ramesh, Tolga Tasdizen + Show 1 more

Open Access

https://doi.org/10.1155/2017/4080874

Copy DOI

Abstract

This paper discusses an algorithm to build a semisupervised learning framework for detecting cells. The cell candidates are represented as extremal regions drawn from a hierarchical image representation. Training a classifier for cell detection using supervised approaches relies on a large amount of training data, which requires a lot of effort and time. We propose a semisupervised approach to reduce this burden. The set of extremal regions is generated using a maximally stable extremal region (MSER) detector. A subset of nonoverlapping regions with high similarity to the cells of interest is selected. Using the tree built from the MSER detector, we develop a novel differentiable unsupervised loss term that enforces the nonoverlapping constraint with the learned function. Our algorithm requires very few examples of cells with simple dot annotations for training. The supervised and unsupervised losses are embedded in a Bayesian framework for probabilistic learning.

Highlights

Automatic cell detection is a fundamental problem that is useful for numerous cell-based studies and quantifications
The maximally stable extremal region (MSER) works well as a region detector only when there is high contrast between the pixels that are inside the cell and those of the background
We proposed a generic semisupervised framework for cell detection that relies on minimal training data

Summary

Introduction

Automatic cell detection is a fundamental problem that is useful for numerous cell-based studies and quantifications. Detection based on local image information can be erroneous, since they can be associated with imaging artifacts or noise. Including prior information about the objects to be segmented helps in resolving these issues. The priors, learned from the training data, can be used to learn strategies to detect cells. Working with minimalistic annotations as in [1], where a dot is placed inside each cell in the training images, may be a more plausible solution. These observations motivate us to study semisupervised methods that use minimal training information

Objectives

Methods

Results

Conclusion