CC BY
10
Научни трудове на Съюза на учените в България-Пловдив, Серия Г. Медицина, фармация и дентална медицина т. XIX. ISSN 1311-9427 юни 2016. Scientific works of the Union of Scientists in Bulgaria-Plovdiv, series G. Medicine, Pharmacy and Dental medicine, Vol. XIX, ISSN 1311-9427 Medicine and Dental medicine June 2016.
КОМПЮТЪРНА ПЛАТФОРМА ЗА ОЦЕНКА НА РЕНТГЕНОВИ
ИЗОБРАЖНИЯ: ПРИЛОЖЕНИЕ ЗА ВАЛИДИРАНЕ НА АНТРОПОМОРФЕН СОФТУЕРЕН МОДЕЛ НА МЛЕЧНА ЖЛЕЗА
Стойко Маринов, Иван Булиев, Живко Близнаков Технически университет - Варна
SOFTWARE PLATFORM FOR EVALUATION OF X-RAY IMAGES: APPLICATION FOR VALIDATION OF ANTHROPOMORPHIC SOFTWARE BREAST MODEL Stoyko Marinov, Ivan Buliev, Zhivko Bliznakov Technical University of Varna
Abstract: Aim of this study is to present a software tool designed to extract texture parameters from x-ray images. The tool is developed in the MATLAB environment and is dedicated to facilitate the research in x-ray imaging. A program called BreastSimulator is used to generate an anthropomorphic software model of an averaged in size breast. Further on, the computer model undergoes simulated mechanical compression to mimic the conditions during mammographic acquisition, and subsequently exported as STL file used for 3D printing. The physical model is printed with a stereolithographic 3D printer with resolution of 100 ^m from Clear Resin. The model is then filled with animal fat and irradiated at GE Senographe digital mammography unit. On the other side, the software model is "irradiated" by a specialized program. The resulting x-ray images from both the simulated and the physical models are compared by their parameters, extracted with the help of the created software tool.
Keywords: x-ray images, texture analysis, software tools
I. Introduction
Anthropomorphic phantoms are important tools used as real tissue and organ substitutes for the purposes of diagnosis and analysis in medicine. The need for such alternatives is implied due to the limitations for the radiation exposure of the human body during x-ray diagnostic procedures, related to the amount of "harmless" radiation, at which there is no risk of causing a cancerous mutation of the irradiated human cells. The anthropomorphic phantoms offer possibilities to design, test and optimize new x-ray imaging modalities and see the effect of the radiation exposure on crucial parameters without causing harmful radiation effect to real human tissues. The anthropomorphic phantoms used in practice split in two main categories - (i) software phantoms, models of human tissues, organs, and parts of them, generated and processed entirely in a computer environment, and (ii) physical phantoms - material objects, representing real tissues and organs, consisting of different materials.
Physical substitutes are being used for longer time than the computer based counterpart, because of the computing power limitations of the technology few decades ago. Presently, these
limitations are overcome by the use of clusters and cloud systems, which allow the accomplishment of very detailed and computationally heavy simulations in real time. Similarly, to the software models, the physical phantoms are used to substitute real human body parts. They are created, either using patient specific data or from computer based models. They are made of different materials, ranging from non-organic polymers and plastics to organic compounds, animal fat and others. These materials must have absorption characteristics close to these of the real tissue. Realistic physical three-dimensional phantoms with a tissue background are needed when investigations on the detectability of lesions, performance of image processing algorithms, reconstruction algorithms, optimization of scanning parameters of new and existing imaging techniques, etc., are required. Manufacturing of such phantoms, however, is associated with several difficulties related to the current technology, suitable materials, manufacturing precision, size of the printed objects, etc. For these new phantoms, a proper evaluation and validation is necessary. Possible approach for that is to extract various parameters from x-ray images obtained from imaged phantoms and compare to the same parameters from clinical images. To facilitate this process, computerized tools are created and used.
Aim of this study is to create a tool, which facilitates the evaluation of such phantoms. The tool computes specific parameters such as standard deviation, skewness and kurtosis, as well as, performs fractal and spectrum analysis of user selected regions of interest (ROI). The developed tool was used in a specific task: to validate a complete model of an x-ray imaging chain, including modeling of a compressed breast, x-ray spectrum and x-ray image formation.
II. Materials and methods
The main purpose of the proposed tool is to provide means for specific processing of x-ray images and data extracting. The initial requirements for the functionality of have been the following:
- Access to images in DICOM format.
- Interaction with the images:
o Performing different computations over the ROIs: (a) Standard deviation; (b) Skewness; (c) Kurtosis; (d) Fractal analysis; (e) Spectrum analysis.
The software tool is created in MATLAB, using the GUIDE module, which allows for the creation of user interfaces as Windows applications. Figure 1 shows a snapshot of the GUI of the developed software tool. It consists of three main subwindows for image importing, region selection and parameter calculation. For each ROI, the following parameters are calculated:
Standard Deviation is a measure that is used to quantify the amount of variation or dispersion of a set of data values. It is calculated by using eq. 1. A low standard deviation indicates that the data points tend to be close to the mean (also called the expected value) of the set, whil e a high standard deviation indicates that the data points are spread out over a wider range of values.
Figure 1. A screenshot from the graphical user interface.
\
1 N
-1 eq. 1
where o is the standard deviation, xi is each value in the data set, |i is the mean of all values in the data set and N is the number of values in the data set.
Skewness it a measure of theasymmetry of the probability distribution of a real-valued random variable about its mean. The skewness value can be positive or negative, or even undefined:
7i = =
f-/i
in ekX-ps]
v
of 1E[(t:-^2]^ Kf'
2 eq. 2
where E isthe e-pect-tion id m the thiod central mnmed g is the sEandard d^e^i^ti-^r^,
I- is ehe mean c^-^11 values ^nitd^ian^'t^ sei.
She Kurtosis measures the relative peakedness or flatness of a distribution. Kurtosis is calculated using eq. 3, where p,4 is the fourth moment about the mean and o is the standard deviation.
Fractal analysis is assessing fractal characteristics of data. For planar images, the 'fractal dimension' ie o chaoactenm tic of ^r mheeert =eoture aa shy eegwns of the image. A lower fractal simauswa coweeponda to tmao^mr textmre end opposite. A mammography image of a fatty brensS hah a ver°t codrte tentont dua do cSet gre-ei contmst dEaahv-en tho porMornttf tintue and lhe phndominated- Wea-i glanctut-ae tioeue, wMle tahe ^e^m^<hat=et]ete^ image of a dense bneast appnaanic Sae- tmoofoee textom sabm eorreepondds ta bwea feactal dimenswn. Thtfrnepod himensioe of a Mage on cniAsetoee? aes
A(s)=Ase-D a
v/ eq. 4
where A(s) is the area of the surface measured with a square of side s, X is a scaling constant
and Disaconstantcharacteristic of the surface (i.e. 'fraetal dimensien').
d'ow-rLao Opscfae A oa/pses: <Пteddower s^ec^mm of the ROIs was calculated by integrating the power spectrum density over concentric rings in the 2D frequency plane using the method in (Bliznakova 2010). Ifehe imugr is marked asp^a,O0, eShn Us power spocPrtm iscomdGted f-om -st dieospte Founter fav^rst:
6f. 5
i M -1N-1
Ff ^T^S Tffc bC-O
MN c=0 b=0
-s 2=r\ — + —
M N
k = 0,1,iiiM -1 I = 0,1,iiiN -1
eq.6
wherek andlare thespatialfrequencies in the two directions, and M x N is the image size off(a,b).
The concentric rings represent an octave sectioning of the frequency plane. The ring width is half the width of the adjacent ring that covers higher frequency information, as the first octave corresponds to the highest frequency ring. The highest frequency is the Nyquist one. The total power spectrum per concentric ring is further transformed to log2(total power spectrum) and plotted versus the octave number. The plot is almost linear and the data are fitted to a line obtained
after linear regression analysis. The slop of this line, m, is related to exponential parameter p as following:
m = p - 2 eq. 7
III. Results
The software tool was applied to validate the accuracy of the
BreastSimulator [1] to generate correctly x-ray mammography images. For this purpose, a small breast phantom was generated (Fig. 2a). Then, the breast model was compressed by using the same software application. The obtained compressed breast was subjected to processing, such as removing all tissues different than gland and skin (Fig. 2b). The model was printed with a stereolithography printer and subsequently filled with animal fat (Fig. 2c).
fr n
Figure 2. Flowchart of the process for obtaining physical breast phantom (a) uncompressed breast model, (b) compressed breast model, (c) physical model.
(a) (b)
A mammography image of the physical phantom was acquired with a GE Senographe SD digital mammography unit featuring a detector with a square pixel size of 0.1 mm and fully automatic exposure. The same geometry and exposure conditions were simulated with the BreastSimulator. Simulated and experimental images are visually compared in Fig. 3. Further on, the calculated parameters for the quantitative comparison are summarized in Table 1.
Table 1. Calculated parameters for simulated and real images.
Parameter Real image Simulated image
P -2.43 ± 0.074 -2.51 ± 0.081
FD 2.14 ± 0.015 2.06 ± 0.001
Skewness 0.48 ± 0.335 0.47 ± 0.300
Kurtosis 1.76 ± 0.347 1.82 ± 0.260
IV. Discussion and Conclusions
The visual comparison between the images in Figure 3 shows a similar appearance of simulated and real mammography image. The objective quantitative evaluation demonstrates very good matching of the parameters for comparison. The graphical user interface is created in a simplistic, minimalistic and a user-friendly way, so that the whole process is visually simplified and easy to use.
This paper presents an in-house developed custom software tool, which facilitates the evaluation of x-ray images by extracting and calculating specific parameters, such as standard deviation, skewness and kurtosis, as well as, fractal and spectrum analysis. With the creation and use of such a specialized tool the time necessary for parameter extraction and comparison decreases significantly and the comparison procedure is facilitated.
References
[1] Bliznakova K, Sechopoulos I, Buliev I, Pallikarakis N, 2012, BreastSimulator: A software platform for breast x-ray imaging research. J Biomed Graph Comput, 2(1), pp. 1-14
