Detection of blob objects in microscopic zebrafish images based on gradient vector diffusion

Gang Li, Tianming Liu, Jingxin Nie, Lei Guo, Jarema Malicki, Andrew Mara, Scott A. Holley, Weiming Xia, Stephen T.C. Wong

Research output: Contribution to journalArticlepeer-review

30 Scopus citations


The zebrafish has become an important vertebrate animal model for the study of developmental biology, functional genomics, and disease mechanisms. It is also being used for drug discovery. Computerized detection of blob objects has been one of the important tasks in quantitative phenotyping of zebrafish. We present a new automated method that is able to detect blob objects, such as nuclei or cells in microscopic zebrafish images. This method is composed of three key steps. The first step is to produce a diffused gradient vector field by a physical elastic deformable model. In the second step, the flux image is computed on the diffused gradient vector field. The third step performs thresholding and nonmaximum suppression based on the flux image. We report the validation and experimental results of this method using zebrafish image datasets from three independent research labs. Both sensitivity and specificity of this method are over 90%. This method is able to differentiate closely juxtaposed or connected blob objects, with high sensitivity and specificity in different situations. It is characterized by a good, consistent performance in blob object detection.

Original languageEnglish (US)
Pages (from-to)835-845
Number of pages11
JournalCytometry Part A
Issue number10
StatePublished - Oct 2007


  • Alzheimer's disease
  • Blob object detection
  • Elastic deformable model
  • Gradient vector diffusion
  • Image analysis
  • Image-based phenotyping
  • Retina
  • Somitogenesis
  • Zebrafish

ASJC Scopus subject areas

  • Hematology
  • Cell Biology
  • Pathology and Forensic Medicine
  • Biophysics
  • Endocrinology


Dive into the research topics of 'Detection of blob objects in microscopic zebrafish images based on gradient vector diffusion'. Together they form a unique fingerprint.

Cite this