Skip to content

Latest commit

 

History

History
215 lines (163 loc) · 5.58 KB

semiautomated_hippocampus_tracing.md

File metadata and controls

215 lines (163 loc) · 5.58 KB
Raw

How to Process MRI Images:

  1. Beginner's Guide to Command-Line Interface
  2. Installation Instructions
  3. Preprocessing T1 images
  4. Hippocampus Segmentation

Reading Materials and Lecture Slides

Set Environment Variables

$ export ANTSPATH=/usr/local/antsbin/bin/
$ PATH=${ANTSPATH}:${PATH}

To permenantly change this, edit your bash_profile:

$ vi ~/.bash_profile

Press the a key to edit text file and add the text above. When you are done editing press the esc key. To save the file, type :w and hit enter. To quit the text editor, type :q and press enter.

Generate a Brain Mask and Skull Stripped Brain

$ c3d <segmentationImage>.nii.gz -binarize -o <outputMask>.nii.gz
$ c3d <inputImage>.nii.gz <maskImage>.nii.gz -multiply -o <outputImage>.nii.gz

Register Template to Participant Image Using Landmarks

Set Parameters

$ FIX=<template>.nii.gz
$ FIXL=<templateLandmarks>.nii.gz
$ FIXM=<templateTracing>.nii.gz
$ MOV=<inputImage>.nii.gz
$ MOVL=<inputLandmarks>.nii.gz
$ OUT=<outputPrefix>
$ ITS=100x100x100x20
$ DIM=3
$ LMWT=0.9
$ INTWT=4
$ PCT=0.8
$ PARZ=100
$ LM=PSE[${FIX},${MOV},$FIXL,$MOVL,${LMWT},${PCT},${PARZ},0,25,10000]
$ INTENSITY=CC[$FIX,${MOV},${INTWT},4]

Run ANTs

$ ANTS \
$DIM \
-o $OUT \
-i $ITS \
-t SyN[0.1] \
-r Gauss[3,0.] \
-m $INTENSITY \
-m $LM

Warp Images

# Warp Participant to Template

WarpImageMultiTransform \
$DIM \
$MOV \
${OUT}ParticipanttoTemplate.nii.gz \
${OUT}Warp.nii.gz \
${OUT}Affine.txt \
-R $FIX

# Warp Template to Participant

$ WarpImageMultiTransform \
$DIM \
$FIX \
${OUT}TemplatetoParticipant.nii.gz \
-i ${OUT}Affine.txt \
${OUT}InverseWarp.nii.gz \
-R $MOV

# Warp Tracing to Participant

$ WarpImageMultiTransform \
$DIM \
$FIXM \
${OUT}auto.nii.gz \
-i ${OUT}Affine.txt \
${OUT}InverseWarp.nii.gz \
-R $MOV

Exclude Everything NOT Gray Matter

$ c3d \
<autoSeg>.nii.gz \
<graymatter>.nii.gz \
-multiply \
-o <outputImage>.nii.gz

Generate Learning Classifiers for Correcting Segmentation Errors

STOP THIS STEP TAKES SEVERAL DAYS ON SOME COMPUTERS!!!

$ bl \
3 \
-ms *<manualSeg>.nii.gz \
-as *<autoSeg>.nii.gz \
-tl <targetLabel> \
<ouputPrefix>

Apply Corrections to Segmentations

$ sa \
<autoSeg>.nii.gz \
<AdaBoost_outputPrefix> \
<outputImage>.nii.gz

ROI Volume Overlap (AKA Dice's Coefficient)

Dice's Coefficient is defined as:

To calculate Dice's Coefficient:

$ c3d \
-verbose
<manualSeg>.nii.gz \
<autoSeg>.nii.gz \
-overlap <targetLabel>

Extract Total Number of Voxels and Calculate Volume

$ fslstats \
<inputSegmentation>.nii.gz \
-l <1-targetLabel> \
-u <1+tagetLabel> \
-V

Useful Code

Save Output Values to Text File

At the end of the c3d -overlap or fslstats code add:

>> <outputFile>.txt

Use TextWrangler.app to edit your file into the standard wide format that can easily be imported into statistic software packages like SPSS, R, SAS, etc. In other words, each row represents data from one entity while each column represents a variable.

Split an ROI into two ROIs

If you somehow get both the left and right hippocampus as a single ROI, there is a way to seperate the two ROIs:

$ val=$(fslval <inputSegmentation>.ni.gz dim1)
$ xsize=$(echo "$val/2" | bc)
$ fslroi <inputSegmentation>.ni.gz <outputLeft>.nii.gz 0 $xsize 0 -1 0 -1
$ xmin=$xsize; xsize=$(echo "$val-$xmin" | bc)
$ fslroi <inputSegmentation>.ni.gz <outputRight>.nii.gz $xmin $xsize 0 -1 0 -1
$ c3d <outputLeft>.nii.gz -binarize -o <outputLeft>.nii.gz
$ c3d <outputRight>.nii.gz -binarize -o <outputRight>.nii.gz
$ fslmaths <outputRight>.nii.gz -mul 2 <outputRight>.nii.gz
$ fslmerge -x <outputSegmentation>.nii.gz <outputRight>.nii.gz <outputLeft>.nii.gz
$ rm <outputRight>.nii.gz
$ rm <outputLeft>.nii.gz

Homework

Due on Monday, October 6, 2014

The dataset for this class located on the ds214 NAS drive now contains these files: n4_resliced.nii.gz, segmentation.nii.gz, t2.nii.gz, brain.nii.gz, HIP_auto.nii.gz, HIP_seg.nii.gz, HIP_manual.nii.gz

Overall does any group show reduced right or left hippocampus volume?

Extract the volumes from the semi-automated hippocampus tracing (HIP_seg.nii.gz). Then, generate a plot comparing the two groups (13* and 23*) left hippocampus and right hippocampus.

Does the hippocampus correlate with overall gray matter or total brain volume?

Next, from the segmentation file extract the gray matter volume and then total brain volume ($\text{gray matter} + \text{white matter}$). Correlate hippocampus volume with just gray matter volume and then total brain volume.

Remember to make your graphs pretty: provide a title, label your x- and y-axis, color code your two groups, include the standard error on bar plots, etc.

Further Readings