Recent Projects

Learning-Based Image Classification Tool: BIOCAT

BIOCAT (BIOimage Classification and Annotation Tool) is a pattern recognition based open framework for automatic annotation and classification of 2D and 3D biological image sets and regions of interest (ROIs) in the images. It enables the comparison and selection of multi-dimensional feature descriptors and learning methods that are suitable for solving given biological image classification problems.

Link to software website:  BIOCAT

Associated papers:
Zhou et al. 2013 (
Zhou et al. 2013 (ACM BCB 2013) (

High-throughput Analysis of Dendritic Metrics

This project is based in obtaining an automatic approach to analyze and quantify dendritic characteristics important for studying neuronal function. Specifically, this approach is critical for making automated fluorescent imaging and screening high-throughput. We developed algorithms and a tool for quantifying dendritic length that is fundamental for analyzing the growth of neural networks (Neuroinformatics 2015).

Link to software website: Dendrite Length Quantifier

Associated Paper:
Neuroinformatics 2015 (

Automated Analysis of High-Dimensional Neuronal Images and Recordings
  •  Automatic quantification of 3D Axonal Terminal Topography
    We develop 3-dimensional (3D) analysis methods to analyze the molecular mechanisms underlying the spatial arrangements of adjacent sensory terminals (i.e., fine-scale topography) of Drosophila neurons (Yang et al. Current Biology, 2014)
  •  Automatic Analysis of Synaptic Distribution in High-dimensional Complex Neuronal Images
    We employ a novel learning-guided system for counting synaptic markers in 3D confocal neuronal images in Drosophila Iobula plate tangential cells (Sanders et al. BMC Bioinformatics, 2015).
  • Dynamic Neuronal Activity and Behavior Analysis
    We investigate algorithms to analyze whole brain neuronal activity patterns.

Associated Papers:
Sanders et al. BMC Bioinformatics, 2015 (

Yang et al. Current Biology, 2014 (


Multi-Class Learning and Classification

We study machine learning algorithms that can effectively decompose a multi-class problem (Pattern Recognition 2008, IEEE DSAA’2015).