Time 
Mon 
Tues 
Wed 
Thurs 
Fri 
8:00  9:30

Intro & Orientation (Fowlkes)
Highdata Rate Microscopy (Gratton/Digman)

Biological Applications
of FCS (Digman)

Spatial And Graph
Data Analysis
(Smyth)

Machine Learning (Xie)

Highperformance
Image Analysis
(Fowlkes)

9:30  10:00

Coffee break 
Coffee break

Coffee break

Coffee break

Coffee break

10:00  12:00

Fluorescence Correlation (Gratton) 
Fluorescence Correlation Lab (Golke)

Spatial and
Graph Analysis Lab

Machine Learning Lab

Highperformance
Image Pipeline Lab

12:00  1:00

Lunch @DBH 
Lunch @ DBH

Lunch @ DBH

Lunch @ DBH

Lunch @ DBH

1:00  2:30

Fluorescence Correlation Data Acquisition (Gratton/Digman)
NatSci II, Rm 3331 & LFD

Image Processing (Fowlkes)

Guest lecture:
Suni Gandhi (UC Irvine)

Detection, Segmentation and Tracking (Fowlkes)

Data Visualization
(Gopi)

2:30  3:00

Coffee break 
Coffee break

Coffee break

Coffee break

Coffee break

3:00  5:00

Fluorescence Correlation Lab contd. 
FCS Data Analysis Lab LFD Computer Lab, NatSci II

Calcium Imaging
TimeSeries Lab

Cell Detection and Tracing Lab

Visualization Lab

5:00  6:00

Debrief 
Debrief

Debrief

Debrief

Course wrap up
& exit survey

A brief overview and introduction of the course topics and our motivations for teaching it.
This lecture provides an overview of available high datarate microscopy systems, basic physical constraints on microscope designs, and approaches for optimizing configurations for the collection of image volumes with high spatial and temporal resolution. Operating principles of modern fluorescence microscopes including optical sectioning, confocal microscopy and cutting edge developments in instrumentation including Selective Plane Illumination (SPIM) and super resolution microscopy.
3. Lecture: Fluorescence Correlation Microscopy [Gratton, Digman]
Single molecular detection is critical for measuring diffusion, binding kinetics and aggregation state of proteins. In a microscope setup, fluorescence fluctuations of tagged proteins are detected in a small focal volume (~1015L). Detection in a very small volume increases molecular number fluctuations and it allows correlation and crosscorrelation of single molecular fluctuations using the autocorrelation function of the fluorescence intensity. These lectures will provide fundamental understanding of the physical properties of molecular motion, flows and transport. It will also specifically cover the fluorescence Diffusion Tensor Imaging (fDTI) analysis. This lecture will describe the basic mathematical and computational tools needed to understand correlation analysis in one, two and three dimensions.
4. Lab: Fluorescence Correlation measurement and Analysis [Gratton, Digman]
Lecture will describe case study of processing big image data produced by very fast acquisition of FCS data and need to optimize the downstream processing. Students will tour LFD and opportunity to collect data with uSPIM instrument. Students will perform hands on computational correlation analysis of FCS datasets.
5. Applications of Fluorescence Correlation Microscopy [Digman]
Live single cell spatiotemporal analysis of protein dynamics provides a mean to observe stochastic biochemical signaling which may lead to better understanding of cancer cell invasion, stem cell differentiation and other fundamental biological processes. This talk will describe an application of the pCF (pair correlation function) analysis to understand p53 activity as an example of protein dynamic interactions in living cells. p53 is a tumor suppressor protein that regulates target genes involved in DNA damage migration and repair. If cells become stressed due to DNA damage, p53 will form tetramers at specific chromatin sites and will activate gene transcription of specific proteins that trigger cell cycle arrest or apoptosis. To gain information regarding fast dynamic processes of p53 behavior, we have imaged p53 with the single plane illumination microscope (SPIM) suing a fast sCMOS camera as a function of time to map different modes of diffusion: confined, directed and Brownian. These analyses can provide answers to the following biological questions: how fast are proteins interacting in their local environment and at what spatial scale?
6. Lab: Highperformance Analysis of Fluorescence Correlation [Gohlke]
Scaling up image processing to very large datasets involves using a variety of techniques to optimize the processing of individual images, splitting computational tasks across multiple processors and optimizing the transfer of image data from storage to memory to cpu. This lab will explore the optimization of a relatively simple correlation calculation in order to speed it up by orders of magnitude and allow interactive visualization of results. Narrative discussion of engineering development of the software, basic approach/software libraries used to make it fast and scalable, analysis of resulting performance speedup.
7. Lecture: Image Processing [Fowlkes]
This will provide general mathematical background on processing of digital images. Students will learn basic principles of sampling and linear filtering with application examples of detecting edges and oriented structures. Theory will provide a context for understanding common image processing methods for biological imaging including deconvolution, image warping and registering and blending multiple volumes acquired during tiled acquisition of large volumes.
8. Lab: Analysis of Fluorescence correlation data [Gratton, Digman]
Students will carry out hands on computational correlation analysis of FCS datasets collected during previous lab session using SimFCS and custom tools to analyze molecular diffusion in biological samples.
Lowlevel quantitative image analysis produces noisy local estimates of, e.g. molecular concentrations, diffusion tensors, etc. regularly sampled over an image volume. Such data can be analyzed in terms of models that assume (for example) the existence of locally smooth flows or spatially varying densities. In this lecture students will be introduced to basic statistical tools that are appropriate for modeling such spatial processes from regularly sampled or sparse measurements. Students will also be introduced to ways in which connectivity data can be represented by networks and graphs and how to extract useful insights from relatively complex graphs, characterizing different aspects of the connectivity, including cluster structure, scale, etc. This lecture will cover the basic statistical modeling assumptions and computations needed to fit such models to data.
10. Lab: Spatial and Graph Data Analysis [Smyth]
Students will experiment with computing spatial statistics of structures extracted from image data and perform graphbased analysis of connectivity data.
11. Guest Lecture: Sunil Gandhi
The brain has the remarkable capacity to rewire its connections and thereby alter its function. In the juvenile brain, the plasticity of neuronal connections mediates the finetuning of a wide range of behaviors from visual perception to language acquisition to social recognition. What mechanism regulates the plasticity of connections in the young brain? How might we manipulate neural circuits to reactivate this juvenile plasticity? In this lecture, I will discuss our research on the transplantation of embryonic inhibitory neurons and our discovery that these cells restore juvenile plasticity to the adult mouse brain. I will also review new discoveries we are making about the organization of the visual system using in vivo twophoton fluorescence microscopy and genetically encoded calcium indicators. The lecture will focus on our experimental results as well as challenges we have encountered in the processing and analysis of large fluorescence imaging data sets. Lastly, I will discuss our recent development of high throughput anatomical techniques to visualize the plasticity of neuronal connections using brain tissue clearing and deeptissue light sheet microscopy.
12. Lab: Analyzing timeseries imaging of neuronal activity [Fowlkes]
Students will analyze timeseries data of neuronal activity in visual system of awake mouse imaged using fast fluorescent calcium indicators. Students will explore strategies for denoising and standardizing data, detecting activation sites, and building graphbased correlation models of activity patterns.
13. Lecture: Machine Learning for Image Analysis [Xie]
Modern computer vision methods make heavy use of algorithms that automatically determine useful image features and tune algorithm parameters to maximize accuracy. This lecture will introduce basic formulations of classification and regression prediction tasks and architectures for performing these tasks based on linear prediction, decision trees and “deep” artificial neural networks. Students will gain a highlevel understanding of how these architectures are automatically optimized from training examples and how image features can be selected that provide predictive power. Students will see several casestudies that use these techniques for specific big image data applications and will be introduced to several software packages that implement these methods and make them accessible to end users who lack programming expertise.
14. Lab: Machine Learning for automated analysis of histopathology [Xie]
Students will have hands on practice with using machine learning frameworks including convolutional neural nets (CNNs) for learning detection and classification from labeled training examples. Students will learn strategies for quantifying accuracy of these methods and diagnosing potential problems in their application to data.
15. Lecture: Detection, Segmentation and Tracking [Fowlkes]
To extract meaningful highlevel descriptions of image content such as 3D location and extent of cells requires processing the output of trained classifiers evaluated over the image to produce discrete lists of points (e.g. cell centers), lines (e.g. traced axons) and planes (e.g. membrane surfaces). This lecture will introduce the basic processing of local image features to identify such geometric structures across a range of scales (from intracellular to whole organisms). Discussion of how these methods scale up to very large datasets.
16. Lab: Neuron Tracing Lab [Fowlkes]
Neuron cellular morphology plays a huge role in the function of neurons as information processing units. This lab will explore methods for processing images to produce quantitative descriptions of neuron morphology and mesoscale circuit connectivity from widefield images of labeled neurons in fixed tissue.
17. Lecture: Highperformance computing for Image Analysis [Fowlkes]
Distributing computation over clusters of commodity computers has been a mainstay of scientific computing since the mid 90s, particularly in computationally intensive engineering and physics simulations. In contrast, infrastructure for data intensive computation (relatively few calculations carried out repeatedly on extremely large quantities of data) is undergoing rapid development made possible by falling storage prices (that allow data to be replicated near compute nodes) and robust, open software platforms backed by industry interest in largescale data analytics. In this lecture students will be introduced to background on cluster computing, principles of distributing computations, and practical guidelines for running jobs on compute clusters and multicore architectures.
18. Lab: Scalable Image Analysis Pipelines [Fowlkes]
This lab will provide an overview of the costs, benefits and challenges of using clusters and cloud computing infrastructure to perform big data image analysis. Students will experiment with running image analysis pipelines as batch jobs on a cluster and in the cloud.
19. Lecture: Big Data Visualization [Gopi]
The capability to interactively visualize very large image data and spatially structured analysis results is widely acknowledged as essential to scientific understanding. Visualization techniques in general are evaluated based on their ability to convey the pattern and characteristics of the data in a truthful and easy to perceive manner. This lecture will first discuss good and bad visualizations of data through examples, using this criterion. Then different visualization techniques for different data types will be discussed along with concrete examples. Further, the concept of color and contrast which are important aspects of visualization will be explained. Given these fundamentals of visualization, we will elaborate on visualizing volumetric data and data types derived from imaging data (tensors, vector fields, etc.). Participants will also engage in discussion of scientific ethics associated with data visualization (i.e. tweaking contrast in paper figures, choosing a colormap that hides noise, etc.).
Students will use general purpose visualization tools and some simple scripting to explore different visualization techniques.