Local Weather




Humidity: 41%
Wind: SSW at 11 mph
Wednesday 61°F / 78°F Sunny
Thursday 62°F / 80°F Sunny
Friday 64°F / 80°F Sunny
Saturday 65°F / 76°F Sunny
Sunday 63°F / 74°F Sunny
Monday 63°F / 74°F Sunny
Tuesday 62°F / 73°F Sunny

The BigDIPA course is designed to cover the complete “vertically integratation" of the full image data to knowledge pipeline and to provide a mix of strategies to help participants:

  • Identify appropriate methods and their limits for BIG DATA image acquisition
  • Define and address the issues of BIG DATA image formats, storage and handling (hardware and software considerations)
  • Define the mathematical and statistical frameworks for BIG DATA image processing
  • Implement computational algorithms on appropriate architectures (e.g. GPUs and parallel high performance computing platforms)
  • Visualize processed data (dimensionality reduction and graphical representations)
  • Address information dissemination issues


The course will explore three major areas where BIG DATA scale and complexity impacts and diverges from traditional image analysis procedures:

  1. High-output image data: Acquisition of biological imaging data sources with very high spatial and/or temporal resolutions, resulting in very large datasets (>TByte) associated with a single sample, e.g. imaging fluorescence dynamics within a single cell at high temporal resolution (<1ms/frame) or imaging a whole brain (<1um/voxel). Consequently data handling, analytical methods and computational architectures require different approaches.
  2. Spatial-temporal statistics and connectivity: Very large high-resolution images of biological samples contain rich geometric and dynamical information about structural characteristics of morphology and connectivity at many scales (from cells to whole organisms). Analyzing such images to extract knowledge about connectivity structure requires statistical tools tailored to spatial and geometric data types which are very distinct from, for example, those used for analysis of big data coming from text mining, genomic sequencing, high-throughput mass spectrometry, health monitors, or environmental sensor networks.
  3. Machine learning and data-driven analysis: Processing pipelines for extracting meaningful high-level data from images and video, such as detecting and counting cells or tracking animal behavior, have traditionally relied on heuristic steps that depend on many hand tuned parameters and thus require significant effort to adapt to new datasets or varying imaging conditions. The BigDIPA course will emphasize the use of modern computer vision tools, based on machine learning, which automatically tune parameters from labeled training examples, allowing greater generalization and robustness.

The course uses image input sources derived from advanced fluorescence microscopy aquisition methods, but the overall course structure and lecture content emphasizes common generalities inherent in big data image processing and therefore will be relevant also to other biomedical input data sources, e.g. biomedical diagnostic sensing (MRI, DTI, Optical Coherence Tomography, doppler OCT), animal behavior (long-running video monitoring, smart vivaria) or environmental monitoring (large-scale aerial, robotic submersible underwater monitoring).


Schedule Overview of BigDIPA Course Activities & Themes

Morning (8:30 am - 1:00 pm)
Afternoon (2:00 pm - 6:00 pm)
A) Topic: theory and discussion
A) Topic: theory and discussion
B) Practical implementation
B) Practical implementation
C) Guided tutorial and exercises
C) Guided tutorial and exercises
Day 1
Introduction to BIG DATA: challenges and applications
Big Data Image Acquisition (ex: uSPIM)
Day 2
Physics & Mathematical Frameworks
Basics of Image Processing
Day 3
Machine Learning Frameworks
Computer Vision
Day 4
Data Mining
Data Visualization
Day 5
High-performance Computing
Dissemination - research access, data warehousing (archiving)
Lecture topics:
  • Introduction and Overview of Big Image Data
  • Fundamentals of Fluorescence Microscopy
  • Physics of Fluorescence Fluctuations
  • Basic Image Processing
  • Machine Learning for Bioimage Analysis
  • Detection, Segmentation and Tracking
  • Statistical Analysis of Spatio-Temporal Data
  • Analysis of Networks and Graph Connectivity
  • Interactive Visualization of Big Image Data
  • Introduction to High Performance Computing for Image Data
  • Platforms for Scalable Computation and Storage
Hands on exercises:
  • Paired Correlation Analysis for Understanding Molecular Flows
  • Quantitative Analysis of Cellular Polarization
  • Training Detectors for Cell Detection and Segmentation
  • Semi-automated Tracing of Neurites
  • Clustering Cell Types by Morphology
  • Graph-based Cortical Connectivity Analysis
  • GPU Accelerated Image Filtering
  • Visualizing Vector and Tensor Fields
  • Distributed Data Analysis on the Computer Cluster

Free Joomla! template by L.THEME