Description of FFT and histogram in medical imaging

Jose Jair Arias Castro

Biomedical Engineering Student
TEC de Monterrey
Monterrey, Mexico
[email protected]
Abstract In this electronic report, we are going to do
image process and identified spectra frequencies through
Fast Fourier Transform (FFT) in two dimension. In
addition, intensity histogram to different images of
different methods. Furthermore, we are going to discuss
about changes in frequencies spectra and intensity.
Index TermsFFT, histogram, imaging, Matlab.

I. INTRODUCTION
Medical imaging represent an important part of techniques for
diagnostic, scientists have discovered many methods for example:
MRI, X-ray, PET, ultrasound and Tomography. Although is
important have knowledge about FFT and histogram principles. In
the following text, we are described briefly about this topic.
X Ray: In the simplest form of X-ray imaging or radiography, a
2D projection (shadow or silhouette) of a 3D body is produced on
film by irradiating the body with X-ray photons. This mode of
imaging is referred to as projection or planar imaging [1].
Magnetic Resonance Imaging (MRI): MRI is based on the
principle of nuclear magnetic resonance, normally, nuclei NMR
active have a radiofrequency perturbation called Larmor
Frequency when a magnetic field is applied. When the
radiofrequency perturbation is stopped, RF signals are emitted
from the body and the nuclei return to their previous state of
equilibrium. This RF signals are detected by a radiofrequency
detector for then make an imaging process of signals. The main
advantage of this method is that can obtain a measure of an axis
with appropriate pulse sequences [1]. Today, MRI procedures are
based on the use of magnetic field gradients that make the
resonance field and, hence, the corresponding Larmor frequency
dependent on the position of the resonant nucleus [2].
Computed Tomography (CT): On the other hand, tomography
have a movement of the x-ray source, this can get a better image
reconstruction of the subject. CT work with the same physical
principle of x ray, but in this case only in axis of interest. In
addition, CT makes image reconstruction from the irradiation of x
ray only in the interest zone, this suppose an advantage and
advance to diagnostic of anatomic anomalies [1].
Ultrasound: Ultrasound waves are pressure waves that propagate
through tissue and are reflected at tissue interfaces [2].
Furthermore, when the wave is reflected echo is produced and this
provide a determinate spatial distribution, then we can do
reconstruction imaging. However, wave properties can affected
the quality of measured and details obtain process, for this reason,
the spatial resolution of Ultrasound is limited [1].
Positron Emission Tomography (PET): principle of positron
emission tomography. A positron emitter stabilizes upon emission
of a positron (e+) and a neutrino (location rd). The positron loses
energy due to scattering events and is finally annihilated by
capturing by an electron ( e )at r a . The annihilation event
produces two -rays (511keV) that are emitted in opposite

directions. The imaging object is surrounded by a ring of

detectors [2]. PET (Positron Emission Tomography) is based in a
process called coincidence detection, this method detected
delivered energy when happened annihilation in positrons.
Whereas the difference between SPECT and PET are the
efficiency of detection process and spatial resolution [1].

II. FAST FOURIER TRANSFORM

Originally the purpose of the FFT was to reduce the time
requirement for large number computation , because FFT is a
discrete Fourier transform algorithm which reduces the
number of computations needed for N points from 2 N 2 to 2
N log N , where log is the base- 2 logarithm. As presented
above (Nyquist rate) FFT transform must be harmonically
related to the sampling frequency to avoid anti- aliasing [].
More often parametric estimation in frequency domain (light
wave) is performed using Fast Fourier Transform (FFT). FFT
is an effective tool for computing the frequency component of
altered and unaltered (baseline) wavelengths of light (Walker,
1996; Bracewell & Bracewell, 1986) [4].
Two-dimensional representation for FFT is:

III. HISTOGRAM
A histogram of an image provides information about the
intensity distribution of pixels in the image. The simplest form
of a histogram is the plot of occurrence of specific gray-level
values of the pixels in the image [7].
Normalization of the observed image intensities is of crucial
importance to explore the disease progression in many clinical
studies. However, images from different scanners or with
different acquisition parameters may have large intensity
variations, which greatly affects the results of image analysis
[5].
IV. METHODS
Matlab Guide is a powerful tool to elaborate user interface,
this tool allow a better visualization and easier imaging
process. We have created a user interface to process different
techniques of medical image. First graphic show the
distribution of bottoms and figures, with these elements users
may choose an image of methods as X-ray, MRI, CT,
Ultrasound and PET. Furthermore, user interface have two
bottoms to select a FFT and Histogram. In addition, users can
interpolated the image and then get a FFT and Histogram.

b)

Fig. 1User interface in Matlab Guide

To elaborated code structure, we have selected three basic

commands in Matlab:
1. Y = fft2(X) returns the two-dimensional discrete
Fourier transform (DFT) of X. The DFT is computed
with a fast Fourier transform (FFT) algorithm. The
result, Y, is the same size as X.[8]
2. imhist(I) calculates the histogram for the
intensity image I and displays a plot of the
histogram. The image type determines the number of
bins in the histogram.[9]
3. B=imresize(A,
[numrows
numcols])
returns image B that has the number of rows and
columns specified by [numrows numcols].
Either numrows or numcols may be NaN, in
which case imresize computes the number of
rows or columns automatically to preserve the image
aspect ratio[].Bicubic interpolation (the default);
the output pixel value is a weighted average of pixels
in the nearest 4-by-4 neighborhood [6].

V. RESULTS
A.
X-ray
In the following graphs is showed a magnitude and phase of
FFT of X-ray image. However, the magnitude and phase are
different when X-ray image have a 1024x1024 size (Figure
2). In addition, when we are compared both graphs a soft
variation of FFT phase are present
a)

Figure 1 a) FFT magnitude and phase of original image b) FFT

magnitude and phase of 1024 image.

Histogram of X-ray image, in this case have more pixels

located in low intensity around of the horizontal axis.

Figure 3 X-ray Histogram.

B.
CT
On the other hand, for this technique, we are applied the same
method, but we obtained similar features in the magnitude.
However, both graphs are different. We cannot discard a
change in the phase of FFT.
a)

Figure 6 a) FFT magnitude and phase of original image b)

FFT magnitude and phase of 1024 image.
MRI image have a similar behavior than CT image. MRI
histogram have an intensity proximate to cero.

Figure 4 a) FFT magnitude and phase of original image b)

FFT magnitude and phase of 1024 image.

Figure 7 MRI histogram

D.
Ultrasound
Ultrasound image have a FFT magnitude and phase amplified
because the amount of pixels are bigger in Figure 2b, for this
reason both graphs are changed.
a)
Figure 5 CT histogram.
In this event, the major part of pixels are in the middle zone of
intensity. Therefore, intensity is close to cero.
C.
MRI
We obtained that distribution of energy of FFT was changed in
both graphs. The following graph are showing a less change in
phase.
a)

b)

Figure 8 a) FFT magnitude and phase of original image b) FFT magnitude

and phase of 1024 image.

b)

Ultrasound histogram are described equal distribution on the

horizontal axis in figure 9.

Figure 3 PET histogram.

VI. DISCUSSION

Figure 9 Ultrasound Histogram

E.
PET
In nuclear medicine, we can obtain a PET image in the
following graph their behavior in the frequency domain is
indicated. We can see that this artifact in the first graph of
magnitude is reduced in the second.
a)

When an image interpolation is performed, many features of

FFT graph may change. However, it is valid when the matrix
of values of the interpolated image is bigger than the array of
values of the original image. Therefore, the distribution of
pixels in the intensity will be higher in a histogram of an
interpolated image. Furthermore, this kind of process
establishes a link when the magnitude and phase of FFT is
different by specific spatial locations in the matrix data.
In addition, this facilitates image processing for better results
in the diagnosis. When optical illusions are present in medical
imaging, also, you can generate a better view of frequency and
phase images for better results.
Although some kind of image acquisition techniques may
have a similar histogram, is necessary to identify more
characteristics that enable a better perspective of their
behavior. Possible causes of error approximate mathematical
algorithms and data quantification. However, the purpose of
image processing is to correct errors through acquisition and
identification of undesired events.

b)
VI. CONCLUSIONS
All frequencies spectra are different when the image is
interpolated. Furthermore, FFT are not equal in these methods
because techniques for get data image have diverse physical
principles.
FFT and Histogram may be the first step for then applied
filters, normalization, and similar image process methods,
because data image have special features in frequency domain.
Figure 2 a) FFT magnitude and phase of original image b)
FFT magnitude and phase of 1024 image.
The histogram indicates that it is an image in the lower
intensity of the horizontal axis, a greater number of pixels is
observed in this type of range

Interpolation is an advantage to generate a low frequency

spectra in FFT. Events can be better described and analyzed
after processing and obtaining relevant data such as the
frequency domain and its magnitude.
REFERENCES

[1] Rangaraj M. Rangayyan. (2005). Biomedical Image

Analysis. United States of America: CRC PressLLC.
[2] Rudin, M. (2013). Molecular Imaging: Basic Principles
and Applications in Biomedical Research. [Hackensack] New
Jersey]: Imperial College Press.
[3] Poon, T., & Banerjee, P. P. (2001). Contemporary Optical
Image Processing with MATLAB. Amsterdam: Elsevier
Science.
[4] Echon, R. M. (2014). Advances in Image Analysis
Research. New York: Nova Science Publishers, Inc.
[5] Xiaofei, S., Lin, S., Yishan, L., Wei, Y., Hongpeng, L.,
Peipeng, L., & ... Defeng, W. (2015). Histogram-based
normalization technique on human brain magnetic resonance
images from different acquisitions. Biomedical Engineering
Online, 14(1), 1-17. doi:10.1186/s12938-015-0064-y

[6] Mathworks.com, (2015). Resize image - MATLAB

imresize.
[Online]
Available
at:
http://www.mathworks.com/help/images/ref/imresize.html
[Accessed 27 Aug. 2015].
[7] Kim, D., Huang, H. K., & Dhawan, A. P. (2008).
Principles and Advanced Methods in Medical Imaging and
Image Analysis. Singapore: World Scientific Publishing
Company.
[8] Mathworks.com, (2015). 2-D fast Fourier transform MATLAB
fft2.
[Online]
Available
at:
http://www.mathworks.com/help/matlab/ref/fft2.html
[Accessed 27 Aug. 2015].
[9] Mathworks.com, (2015). Histogram of image data MATLAB
imhist.
[online]
Available
at:
http://www.mathworks.com/help/images/ref/imhist.html
[Accessed 27 Aug. 2015].