Image Processing for Early Diagnosis of Breast Cancer Using Infrared Images

Pragati Kapoor, Dr. S.V.A.V. Prasad

Abstract—Early diagnosis of Breast Cancer is the key to approach using image segmentation and Hough Transform
improve survival rate. Infrared thermal imaging or for the analysis of asymmetry is proposed in this paper.
thermography is a promising screening tool as it is able to
warn women of breast cancer ten years in advance. However, II. IMAGE ACQUISITION
interpretation of a thermogram can be inconsistent. In order to Fig 1, Fig 2 and Fig 3 show the thermograms of a
improve the accuracy of preliminary breast cancer screening
volunteer with a normal mammogram but a concerned
using thermogram, image segmentation is proposed as an
automatic approach for analysis of infrared thermal images.
thickening of the left breast. The images are taken using
Edge detection and Hough transform are outlined for IRI4040 long range thermal imager having a spectral
asymmetry analysis of heat patterns in contralateral breasts response of 8Pm to 14Pm. Fig. 1 shows a significant
increase in temperature of the entire left breast along with
Keywords- Breast Cancer; infrared thermal imaging; noticeable vascularity (angiogenesis)
thermogram; image segmentation; edge detection; Hough
transform; asymmetry analysis

I. INTRODUCTION
Breast cancer has been known for decades to be the most
common type of cancer among women. The incidence of
breast cancer in India is on the rise and is rapidly becoming
the number one cancer in females. India accounts for nearly
six percent of deaths due to breast cancer in the world. One
out of every 22 women in India is diagnosed with breast
cancer [1]. Recent studies have determined that the key to
breast cancer survival rests upon its earliest detection
possible.
However, mammography the most widely employed Figure 1. Front Position
detection method is not as effective for women with
fibrocystic or dense or surgically implanted breasts. In Figure 2, the right breast is seen to be normal and cool
Furthermore, there is concern regarding the risk of ionizing without evidence of suspicious blood vessel activity. Figure
radiation and patients complain of discomfort due to high 3 shows in full detail the significant amount of
compression of breasts. In a search for other imaging thermovascular activity in the left breast. The thermograms
techniques, thermography has emerged as a potential method of Fig. 1and Fig. 3 are risk indicators and it is recommended
to complement mammography and improve overall detection for the patient to have routine follow ups.
efficiency [2]. Thermography possesses many advantages,
such as noninvasive, non-contact, risk free, non-radiation,
less expensive and particularly valuable for early tumor
detection [3]. Ng E.Y.K. et al also mentioned that the result
of thermography can be correct 8-10 years before
mammography can detect a mass [4].
Clinical interpretation of breast thermograms can be done
by making comparisons between the images of contralateral
breasts. The higher chemical and blood vessel activity of the
cancerous tissue will cause an increase in regional surface
temperature of the breast [5]. When the images are relatively
symmetrical, small asymmetries in the heat patterns of
contralateral breasts indicate a suspicious region. These
asymmetric heat patterns are found visually using new
generation of IR cameras. In order to provide a more
objective result, based on the assessment of thermograms, an Figure 2. Right semi side position

978-1-4244-5586-7/10/$26.00 C 2010 IEEE

564 Volumn 3
 Edge Hough
Detection Transform
Original Edge Feature
Image Image Curves

Pixel Pattern
Analysis Classification
Asymmetry Segmented
Identification Image
Figure 3. Left semi side position
Clusters
Skin surface temperature is greatly affected by numerous
conditions. In order to reduce the errors due to thermal Figure 4. Block Diagram for asymmetry analysis of thermogram using
artifacts, the above images are taken using a recommended segmentation
set of instructions to ensure the usefulness and consistency of
thermal images[6]. The patient was asked to avoid alcohol, connected to strong edges. As a result , it is robust to noise
caffeine and stop smoking two hours before the test and and implements Gaussian function to smooth the image and
avoid use of lotion, cream on the body area to be imaged. to obtain the magnitude and orientation of the gradient for
The chest area was cooled slightly with a fan for each pixel.
approximately 7-10 minutes just prior to image taking. The C. Hough Transform for detection of breast boundaries
room temperature was approximately 22 degrees Celsius and
darkened during the test to minimize infrared source To find the lower boundaries of the breast is difficult
interferences. because they are mostly like parabolic curves. Therefore
Hough transform can be applied to extract the boundaries of
III. APPROACH the breasts. Hough transform that was proposed by Paul
Hough is a method to implement a kind of mapping relation
Early methods for interpretation of breast thermogram from image space to the parameter space. The main idea can
were solely based on subjective criteria. The images were be explained as the duality between point and line. The
read for variations in vascular patterning with no regard to points that in the same line in image space correspond to the
temperature variations between the breasts [7]. In this paper, intersecting lines in the parameter space. On the other hand,
image segmentation and asymmetry analysis are proposed as all the lines that intersect at a same point in the parameter
an efficient method for breast analysis using infrared images. space correspond to the points in the same line in image
The approach outlined includes the following steps: space. Hough transform is usually used to find the line or
1) Edge Detection to extract the boundaries of the curve in binary map and transform the binary map into the
breasts. Hough parameter calculation space [9].
2) Hough transform to extract the lower breast An approach based on the Hough transform is as follows:
boundaries. x Complete the gradient of an image and threshold it
3) Classify each segmented pixel into a certain number to obtain a binary image.
of clusters.
x Specify subdivisions in the pT-plane
4) Diagnose the breast diseases based on asymmetric
analyzing of the pixels in every cluster. x Examine the counts of the accumulator cells for high
Fig. 4 is a system guideline of the steps involved in the pixel concentrations.
proposed approach.
x Examine the relationship (principally for continuity)
B. Edge Detection between pixels in a chosen cell.
Edge detection is a well developed field on its own
within image processing. It is by far the most common D. Feature extraction for symmetry analysis of
approach for detecting meaningful discontinuities. contralateral breasts
The built in Canny edge detector in the Matlab Image The features of the heat patterns will include the
Processing toolbox can be used to extract the boundaries of statistical parameters such as skewness, temperature
the breasts because it is one of the most precise edge variation and kurtosis. More larger the tumor size is, the
detection operators and has been widely used [8]. The Canny steeper the temperature variation is [10]. The cumulative
method uses two thresholds to detect strong and weak edges. histogram which describes the temperature variation with
It includes the weak edges in the output only if they are

565 Volumn 3
area will reflect the asymmetry. The histogram will describe [4] E.Y.K. Ng, L.N. Ung, et al., “Statistical analysis of healthy and
the frequency of existence of pixels of the same intensity in malignant breast thermography,” Journal of medical engineering and
technology, vol. 25, pp. 253-263, Nov/Dec 2001.
the whole image or the region of interest (ROI). The
[5] R.D. Leek, “The prognostic role of angiogenesis in breast cancer,”
asymmetry can be measured by the maximum ratio between Anticancer research, Vol 21, no. 6B, pp.4325-31,2001.
temperature difference and area. The further the ratio from 1, [6] E.F.J. Ring and K. Ammer, “The technique of infrared imagingin
the greater asymmetry the is. medicine,” Thermology International , vol. 10, pp. 7-14, 2000.
Skewness and kurtosis can be measured using the [7] M. Gautherie, A. Kotewicz and P. Gueblez, “Accurate and objective
Statistics toolbox in Matlab. Skewness is the measure of evaluation of breast thermograms : basic principles and new advances
asymmetry of data around the sample mean. The Skewness with special reference to an improved computer-assisted scoring
(y) of a distribution x is given by (1). system,” Thermal assessment of breast health, MTP Press Limited, pp.
72-97, 1983.
[8] J. Canny, “A Computational approach to edge detection,” IEEE
E x  P 
3
Trans. Pattern Anal. and Machine Intell., vol. 6, pp. 679-698, 1995.
 y  [9] Cauchie Julien, Fiolet Valerie, Villers and Didier, “Optimization of
V3 an Hough transform algorithm for the search of a center”, Pattern
Recognition [J], vol. 41,. pp. 567-574 , February 2008.
P is the mean of x, V is the standard deviation of x and [10] N. M. Sudarshan and E.Y.K. Ng, “Surface temperature distribution
E(t) is the expected value of quantity x. of a breast with/without tumor,” Int. J. Comput. Meth. Biomechanics
Biomed. Engng. vol. 2, pp. 187-199, 1999.
Kurtosis is a measure of how outlier prone a distribution
is. The kurtosis(k) of a distribution x is given by (2).

E x  P 
4
 k 
V4
The kurtosis of the normal distribution is 3. Distribution
that are more outlier prone than the normal distribution have
kurtosis greater than 3, distribution that are less outlier prone
have kurtosis less than 3.
IV. CONCLUSION
This paper proposes an approach to diagnose breast
cancer based on infrared images. Firstly, the breast picture
can be segmented accurately and automatically with the
application of the methods of edge detection and Hough
transform. Then based on pattern classification and pixel
distribution, asymmetries are identified for the breasts. The
pathological changes of the breasts will be diagnosed based
on this. The approach will be effectual and feasible and
would be of great practical value in diagnosing the
asymmetric abnormalities for breast using infrared images
and will help as a useful second opinion.
ACKNOWLEDGEMENT
The authors thank Irisys for providing high resolution
thermal imagers for taking the breast thermograms and Dr.
Seema Patni, Incharge Breast Clinic, Sunder Lal Jain
Hospital, Delhi for providing expertise for breast cancer
research
REFERENCES
[1] Sherring and Varsha, “Mediating Breast cancer in India,” NCA 94th
annual convention, San Diego, CA, May 2009.
[2] M. Frize, C. Herry and R. Roberge, “Processing of thermal images to
detect breast cancer: A compariosn,” in Proc. 2nd Joint IEEE
EMBS/BMES conf., Houston, TX, pp. 1159-1160,2002.
[3] B.F. Jones, “A reappraisal of the use of infrared thermal image
analysis in medicine,” IEEE Trans. Med. Imaging, vol. 17, pp. 1019-
1027, Dec 1998.

566 Volumn 3