You are on page 1of 20

Lecture 03:

Image Formats & Software


BME 234: Neuroimaging Data Analysis, Spring 2016, Course Code 14310

Frithjof Kruggel, M.D.


Department of Biomedical Engineering, University of California, Irvine
Office: REC 204, Phone: 4-3729, Email: fkruggel@uci.edu

First Prev Next Last Go Back Full Screen Close Quit

Image Understanding System


The process of understanding images is divided into the following steps:

Image formation: Acquisition of imaging data;


Image preprocessing: Standardizing acquired images;
Image segmentation: Breaking up an image into interesting regions;
Object description: Describing imaged objects by quantitative parameters;

Object recognition: Attaching labels to objects.


The intend is to automize processes otherwise performed by humans.

First Prev Next Last Go Back Full Screen Close Quit

Recognition of finger prints


Now, let us examine different image modalities. Consider the aim of image
processing on these data, and the steps to achieve them.
An early application of image processing is the comparison of newly collected finger prints with those stored in a database, e.g. for subject identification. Examples may be found here:
ftp://ftp.cs.columbia.edu/jpeg/other/uuencoded/

First Prev Next Last Go Back Full Screen Close Quit

Recognition of malignant melanoma


Melanoma are skin lesions. Certain visual criteria allow distinguishing malignant melanoma that have to be treated. Recent reognition systems achieve
expert performance. Examples may be found here:
http://matrix.ucdavis.edu/tumors/new/tutorial-intro.html

First Prev Next Last Go Back Full Screen Close Quit

Images and Meta-Data


The documentation of an imaging examination includes:

Image data in 1D (signal), 2D (image), 3D (volumetric image), 3D


(video), 4D (volumetric time series) format;

Image representation information (compression, lossy, loss-less);


Subject-related meta data: Subject name, birth date, exam date and time;
Acquistion-related meta data: Device, modality, protocol parameters;
Context-related meta data: Time point in series, stimulation parameters.
Note that one acquisition may correspond to one or more files. Meta data
may be included in the image or stored separately.

First Prev Next Last Go Back Full Screen Close Quit

File Formats - DICOM


Most typically, acquired data are stored in a proprietary (vendor-defined)
format. Historically, vendors were reluctant to open up data formats,
discouraging inter- operability with other devices. A good overview
on medical imaging formats is found in (Medical Image Format FAQ:
http://www.dclunie.com).
This unfortunate situation in the clinical environment led to the proposal of a
common data format and information interchange protocol called DICOM.
DICOM-capable devices include image sources (MR scanners, US, CT, digital X-ray devices), storage (image servers), viewing stations (consoles), and
film printers.
Vendors are required to publish a DICOM conformance statement that
describes the scope of (and any deviations from) the implementation of the
DICOM standard.
Remember that MR scanning produces data in k-space (Fourier space). Acquired data are reconstructed as an image using an inverse Fourier transform.
Most analysis procedures use images, rarely raw data are required for analysis.

First Prev Next Last Go Back Full Screen Close Quit

DICOM - Overview

First Prev Next Last Go Back Full Screen Close Quit

DICOM - Tags

First Prev Next Last Go Back Full Screen Close Quit

DICOM - Messages

First Prev Next Last Go Back Full Screen Close Quit

DICOM - File Structure

First Prev Next Last Go Back Full Screen Close Quit

DICOM - Discussion
Advantages of this format are: .

standardized, contains tags (patient name, acquisition information etc.);


widely used in hospitals, storage servers.
Several issues remain:

Complex standard with several revisions;


Vendors have proprietary tags and information - some information is not
disclosed;

Only 2D format defined (i.e., volume data are delivered as a series of


slices).

Conversion software is complex, does not cover all DICOM-capable devices, and is often incomplete.

First Prev Next Last Go Back Full Screen Close Quit

File Formats - NIFTI


Neuroimaging experiments typically generate 3D or 4D data that often map
to several thousand DICOM files, with considerable overhead due to repeated metadata. This lead the Neuroimaging Informatics Technology Initiative (NIFTI, http://nifti.nimh.nih.gov/nifti-1) to propose a new data format
for the exchange of neuroimaging data. It builds on an older, still widely
used Analyze format (www.mayo.edu/bir/PDF/ANALYZE75.pdf).
An Analyze data set is comprised of two files, a data file and a header file.
The header file contains all information required to interpret the data file,
which is just a matrix.
NIFTI packs header and data into a single file, and offers loss-less compression. The NIFTI format is a quasi-standard in neuroimaging data analysis.
All packages offer import- and export modules to NIFTI format.

First Prev Next Last Go Back Full Screen Close Quit

NIFTI 1

First Prev Next Last Go Back Full Screen Close Quit

NIFTI 2

First Prev Next Last Go Back Full Screen Close Quit

NIFTI 3

First Prev Next Last Go Back Full Screen Close Quit

NIFTI 4

First Prev Next Last Go Back Full Screen Close Quit

NIFTI 5

First Prev Next Last Go Back Full Screen Close Quit

NIFTI - Discussion
Advantages of this format are:

allows the efficient storage of 2D, 3D, 4D data;


includes most commonly required meta data;
relatively simple format with stable software support;
is a quasi-standard in neuroimaging.
Several issues remain:

not formally standardized;


meta data is not extendible;
some information must be held in external files (e.g., gradient directions);

no representation for signals;


also provides a format for surfaces (GIFTI) that is not widely used;
First Prev Next Last Go Back Full Screen Close Quit

BRIAN - Discussion
BRIAN (BRain Image Analysis) - processing package and image format
developed at MPI and UCI. Based on an image format developed in computer vision, called Vista.
Advantages of this format are:

allows the efficient storage of 2D, 3D, 4D data;


user-extendible meta data format (key-value pairs);
represents images, signals, surface and volumetric meshes;
offers transparent loss-less compression;
allows a set of different data (e.g., a volume and a surface) to be stored
in a single file;

supported by stable and mature software.


Several issues remain:

not formally standardized;


not widely used.
First Prev Next Last Go Back Full Screen Close Quit

Software for Neuroimage Data Processing


SPM Homepage:

http://www.fil.ion.ucl.ac.uk/spm/

strength:

fMRI/PET imaging;

FreeSurfer Homepage: http://surfer.nmr.mgh.harvard.edu/ - strength:


anatomical imaging;

FSL Homepage:

http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FslOverview strength: segmentation;

Brainvisa Homepage: http://brainvisa.info/ - strength: anatomical imaging;

BrainLab Company Homepage: http://www.brainlab.com - strength:


neurosurgical applications;

BRIAN: http://sip.eng.uci.edu/ - strength: anatomical imaging, segmentation, FE analysis.


All software is freely available on these web sites, but may need some effort
to install, understand and use.
First Prev Next Last Go Back Full Screen Close Quit