%* Last edited: Mar 31 15:02 1998 (sw)
\documentstyle[12pt]{article}

\include{new-macros}
\easypage

\begin{document}

\title{Laboratory Project 6 \\Post Processing MRI Images}

\maketitle



% ---------------------------------------------------------------------------
\section{Introduction}

This lab will explore some of the common techniques that are used for
post-processing MRI images.

Each of the following sections will refer to supplied Matlab code
that is located in the following directory ``...''.  In addition,
some clinical MRI images are located in ``...''.  

The data consists of several series located in subdirectories
``001'', ``002'' etc.







% ---------------------------------------------------------------------------
\section{Getting Started}

\subsection{Supplied Code}

\begin{itemize}
	\item ge\_read.m
	\item ge\_write.m
	\item ge\_display.m
\end{itemize}

\subsection{Preliminary Exercises}

Use the above Matlab functions to read and display selected slices
from series 001.

\subsection{Include in Lab Report}

Include in your lab report a picture of one of the slices.

% \subsection{Helpful MATLAB functions}

% ---------------------------------------------------------------------------
\section{Pre-Processing}

\subsection{Supplied code}

\subsection{Thresholding}

Write a function to threshold the data and display the results
of the thesholding. 

\subsubsection{Include in Lab Report}

In the following exercise, use slice foo from series boo.

Separate tissue from air in the data by thresholding.  
Try to separate white matter and gray matter by thresholding.

Turn in images of the results.

Discuss the quality of the results.

\subsubsection{Helpful MATLAB functions}

matlab blurer	: filter2


\subsection{Noise Reduction}

Notice that the segmentation results from above contain
some fragmentation due to contamination by white noise.

A standard method for noise reduction is linear filtering.  
Try to reduce the noise-induced fragmentation by use
of a low-pass filter.

Non-linear noise reduction methods are often the choice
for MRI.  Try the noise reduction again using median
filtering.

Try various settings for the filter parameters.

\subsubsection{Include in Lab Report}

Repeat the threshold segmentation of white/gray matter on 
the filtered images (lowpass filtered
and median filtered).  Include results on slice bazz showing
what you think are the best improvements in segmentation
results.

\subsection{Helpful MATLAB functions}

\subsection{Shading in MRI}

In your attempts to segment white matter and gray matter
above, you may have noticed artifacts (variations in the
apparent thickness of gray matter) due to slow spatial
variations in the signal intensities (``shading'').  

Attempt to correct for this artifact by using
the method of ``unsharp masking''.  

\subsubsection{Include in Lab Report}

Repeat the threshold segmentation of white/gray matter on the
intensity-corrected images. Include results on slice bazz showing what
you think are the best improvements in segmentation results.

\subsection{Helpful MATLAB functions}


% ---------------------------------------------------------------------------
\section{Statistical Classification}

This exercise will involve the construction of a statistical
classifier for MRI.  We will focus on segmenting the MRI slice data in
.  Training data for this slice is available in ... .  This image-like
file consists of small integers that label the tissue type at each
pixel.  The types are encoded as follows.. .


\subsection{Supplied Code}

\subsection{Preliminary Exercises}

As a first step, construct histograms of the training data.

Fit Gaussian densities to the training data for each tissue type.

Finally, classify each pixel in the MRI data for slices
XXX and YYY using the Gaussian tissue
class conditional intensity models constructed above.  Construct a
result segmentation where the tissue at each pixel is represented
using the encoding scheme described above.


\subsection{Turn In}

Include in your lab report histograms for each tissue type
in the training data.  Display the Gaussian class
conditional intensity models by plotting the density
functions over their corresponding data histograms.


Prepare and include images displaying the training segmentation and
your classification for slices XXX and YYY.  In these images, display
each of the tissue classes with a different intensity.
Discuss...

\subsection{Helpful MATLAB functions}


% ---------------------------------------------------------------------------
\section{Post-Processing}

This exercise will explore simple methods for isolating
an anatomical structure, the brain, from MRI.

\subsection{Supplied Code}

\subsection{Preliminary Exercises}

Working with the training segmentation, ..., construct
a binary image that represents brain and non-brain.
Notice that due to contrast failures and noise,
there are numerous spurious bits of 
``brain'' located outside of the ``real'' brain.

Write a Matlab function to find the largest connected
component of ``brain''.  Notice that this removes the
isolated errors, but there are still some spurious
structure connected to the isolated brain.

Morphological methods (erosion, dilation, and especially opening)
are frequently used to remove thin structures from a bulkier
structure.  Try using opening (erosion followed by dilation)
to trim the ``junk'' that is stuck to the brain.

\subsection{Turn In}

Include in your lab report images of the isolated
brains before and after the morpological processing.

Discuss...


\subsection{Helpful MATLAB functions}

% ---------------------------------------------------------------------------
\section{Validation}

\subsection{Supplied Code}

\subsection{Preliminary Exercises}

Design some measures for comparing your segmentations
against the atlas.  

\subsection{Turn In}

Compare your segmentations against the atlas using
the measures you came up with.  Do this for the
results before and after the morphological processing.

\subsection{Helpful MATLAB functions}

% ---------------------------------------------------------------------------
\section{Visualization}

\subsection{Supplied Code}

\subsection{Preliminary Exercises}

\subsection{Turn In}

\subsection{Helpful MATLAB functions}


\end{document}