You are on page 1of 4

1) This study developed an automatic kidney stone

Automated Size Measurements


measurementsofprogram
Kidney Stone
that integrated digital image
processing and image analysis techniques in CT images.
on 3D CT Image
Lai Yee Myint1, Khine Thin Zar2
1
Department of Information Technology Engineering
West Yangon Technological University
2
Department of Computer Engineering and Information Technology
Yangon Technological University

1leiyee2013@gmail.com
2khinetz2021@gmail.com

Abstract - Kidney stone is a common disease with an


increasing incidence worldwide. Computed Tomography Specifically, the study (1) developed a method for
(CT) value is the essential diagnostic parameters for the selecting the largest slice of the whole 3D object (2)
management of kidney stone treatment as a less laborious developed a method for calculating the stone
or expensive method. The goal of this study is to develop measurements such as area, average HU, maximum HU,
the automated measurement of area, mean Hu, maximum minimum HU and density.
Hu, minimum HU and diameter from chest CT scans on
diagnosis of kidney stones as compared to manual 2-D
measurements. In the established three-dimensional
II. LITERATURE REVIEW
reconstruction model, the proposed study was measured CT is the gold standard and its CT values are useful
the CT measurements such as the area, mean of HU for the pre-operative study of stones and influences the
(Hounsfield units), the maximum and minimum HU value, choice of surgical strategy. In [3], the program was
the diameter and the density. It is an automated system develop to measure the stone volume by using two 2D
that can select the slice with the largest area among the
measurements, maximum stone length and surface area.
whole slice and calculated the CT value. According the
results, the present study can provide the more precise CT Although this was the optimal method for assessing
values on kidney stone CT as compared to manual stone volume and stone burden, there were some
estimation. Moreover, it can reduce the processing time limitations for the stone with irregular shape and
and can measure the CT values not depend on technician. staghorn calculi. In [4], HU density (HUD) was defined
Therefore, the proposed system for kidney stone diagnosis based on the mean HU and the largest diameter of the
is very effective to select the largest slice with the program stone. Although CT value of this study could help to
and identify stone CT values without any manual process. predict stone composition and residual fragments after
PCNL, it is impossible to develop with the help of
Keywords – CT measurements, area, largest diameter, urologist. In [5], skin-to-stone distance, HU density,
mean Hu, maximum Hu and minimum HU. axial stone diameter and stone volume were calculated to
produce stone free status after extracorporeal shock
I. INTRODUCTION wave lithotripsy (ESWL). This study is an optimal
Kidney stone disease is the risk of life in throughout the predictor of SF status after ESWL. Nevertheless, it could
world, and majority people with initial stone formation determine the maximum stone length only if there were
impossible to notice it as disease. Later, it can damage three CT types, CT in coronal, sagittal and axial planes.
the kidney slowly. Many people affected by continual The SSD was also calculated with the help of
kidney failure before viewing symptoms, and until radiographic callipers. In the knowledge of our study,
kidney death. Although stone formation can be quite there is still challenges to develop the CT values without
complex with various stone composition, most of stone the help of manual process.
are growing in the renal calyces. They are leading to the
ureter after passaging into the ureter. Since kidney III. METHODOLOGY
failure can be life threatening illness, early diagnosis of From No_2, Defence Services General Hospital,
diseases is crucial [1]. The currently available imaging Yangon, Myanmar, CT images of kidney stone cases
modalities include CT image which is widely used were obtained by using a Toshiba Aquilion 64-slice CT
imaging techniques for diagnosing kidney diseases. CT Scanner. This set of CT scan is used as the input data of
with high specificity demonstrate that a patient’s the proposed system. The study group included both men
symptoms are related to stones when they are visualized and women with any kidney stone cases. All scans were
[2]. Stone burden has traditionally been assessed by obtained with an intermediate dose CT protocol for the
measuring the maximum length, area and mean HU on urinary tract (tube voltage: 120 kV, Current Exposure
non-contrast-enhanced CT. As kidney stones are time: 80-130 mA, 0.5s scan time (rotation time), 380
irregular three-dimensional (3D) structures that often width and 40 length, Focal spot (Mag) 1.4x, display field
have complex geometric shapes, manual two- of view (DFOV) 33.1 × 33.1 cm). The weight of the
dimensional (2D) measurements may inaccurately included patients is unknown and the noise (stents and
predict thus not reflect total stone burden with long calcification) depends on the patient since no record was
processing time. Accurate stone measurements are of presented. A solitary ureteral stone is present in various
significant clinical importance for kidney stone location and size. The existence of stone (present or
diagnosis and treatment decision [3]. absent) been provided by a radiologist. The full dataset
consists of 75 clinically acquired unenhanced abdominal
CT scans data with 0.705×0.705×1 mm voxel size and 1
mm slice spacing. Each patient consists of the number of
slices between 500 and 600. Every cross-sectional slice
is a two-dimensional image of 512 ×512 matrix size and
1 mm slice thickness, these are called the x-, y- and z-
dimensions respectively. CT image is 3D gray scale
image where the position of each pixel represents the
intensity of the gray level.
The program design was grounded on the application
of image processing techniques and geometry principles.
Image processing techniques applied in the program
include Binarization [4], morphological operation [5], Figure 2. Result of the stone area when applied to a
localization (bounding box) [6], Boolean operation [7] slice
and connected component [6]. The study is done step by B. Measuring the stone area using logical OR operation
step as shown in Figure 1. Beginning with binarization result of the first slice
(and moving progressively to the last slice), all pixels
equal to 1 (foreground) were counted and other pixels
equal to 0 were established. Then, the total number of
pixels at foreground are outputted as the area of stone at
each slice.

C. Selecting the largest slice using Fuzzy Logic (If-then


rule)
Fuzzy logic and bounding box are applied to select
the largest slice of a stone. To overcome computation of
distances around the stone when applying connected
component, the program establishes a bounding box
from individual computation of the distance from the
topmost and bottommost pixel (length, x), and the
rightmost and leftmost pixel (width, y) of the abdomen
to generate two horizontal and vertical lines that will
enclose the area. And then, the number of slices of the
stone is regarded as the height (z) of a stone. The slice
with the largest diameter of each object was selected as
follow,
Step 1: Set the slice, Si=(S1, S2, S3, …, Si)
Step 2: Set the area of each slice, ai=(a1, a2, a3, …, ai)
Step 3: Larea=0, Lslice=1;
For Si=1:i
If (Larea<Si(ai))
Lslice=i
End
Larea=ai
End
According the above algorithm, the largest slice of a
stone is selected.
Figure 1. The block diagram of the proposed program
D. Measuring the CT values
A. Localization CT values are calculated on the slice with the largest
It was applied to slices where the stone is visible to be diameter of each stone.
largest area. Scanned images generally include extra
parts around stone in Figure 2. Binarization was applied i) Area: Area of a kidney stone at each slice is
in each slice to allow segmentation of the foreground specified by counting the number of pixels in the specific
(stone) by a white colour and the background by a black region.
colour. Morphological operation such as dilation and
erosion were applied respectively to isolate the object of
interest in the foreground. This allows detection of the
connected component through application of the 8- ii) Average Hounsfield Units: For CT, intensity
connection connectivity which identifies any component represented the Hounsfield units (HU) in each pixel.
pixel equal to 1 in the foreground. All pixels that equal Average Hu of each stone is the ratio of total HU
to 1 are connected; if a pixel is equal to 0 a boundary of value of all pixel in the slice into the total number of
the connected component will be established. pixels at that slice. Average HU of each stone is
calculated as follow:
referenced with the manual records through the
physician.
A. Manual Data Analysis
Manual two-dimensional CT measurements may be
Where, is average HU for each stone, n is the inaccurate because it is completely depended on the
number of pixels in a slice. µ is the HU value of each technician. It is impossible the precise result by
pixel in a slice. measuring the different technicians. The results can vary
by choosing different slices. The manual record of this
iii) Maximum Hounsfield Units: Maximum Hu of an study could provide the estimate CT values of the stone.
object is the maximum Hu value of a voxel in an object. In Figure 3, the region of interest (ROI) of manual
It is calculated the following algorithm. record was smaller than the actual stone area and then
Step 1. Set HU=( , ) , is other values were also calculated on this ROI.
represented the HU value of voxel in an object.
Step 2. Maximum HU value,
Step 3. For i= 2 to i,
If >
=
End
End
Step 4. Solution=
Minimum Hounsfield Units
iv) Minimum Hounsfield Units: Minimum Hu of an
object is the maximum Hu value of a voxel in an object.
It is calculated the following algorithm.
Step 1. Set HU=( , ) , is
represented the HU value of voxel in an object.
Step 2. Minimum HU value,
Step 3. For i= 2 to i,
If <
=
End
End
Step 4. Solution=
Figure 3. CT values (a) area, mean HU, maximum HU
v) Diameter: The diameter is one of the essential CT and minimum HU (b) largest diameter of manual record
values. It was also calculated on the largest area slice of of a stone
a stone as the following algorithm. B. Performance Evaluation
Step 1, Set bounding box “R” on the stone area of the The results (area, average HU, minimum HU,
slice. maximum Hu, diameter and density) of the study among
Step 2, Calculated x and y, the number of pixels of the patients clinically diagnosed with program (Group I) and
length and width of the bounding box, R. the manual results without program (Group II) are
Step 3, Calculated the largest diameter “D” by choosing compared to evaluate the proposed study as shown in the
the larger one from x and y. Table 1.
HU Density
According to the analysis, most of the result of the
vi) HU Density (HUO): HU Density (HUD) of the stone proposed program are approximate with the manual
is the ratio of the average HU value of the largest slice of record. In circular shape stone (stone 11, 14 and 15), the
the stone into the diameter of that stone. program is able to output almost exactly results. Since
most of stone are irregular shape and unsmooth
boundary, the manual record is impossible to output the
exact results. Therefore, the proposed program can used
to measure any shape of stone. Moreover, the results are
Where, HU Density is the density of a stone, HUavg is
always the same for the measure of anyone. It can
average Hu for each stone and D is the largest diameter
overcome the trade-off between the time and the
of the stone.
accuracy of manual processing.
IV. RESULT AND DISCUSSION
.
Data analysis for this study was based on the
incidence of kidney stone detected among patients.
Results generated by the proposed program were cross-
TABLE I
CT VALUES BY THE PROPOSED STUDY AND MANUAL RECORD
Stone Area Average HU Maximum Hu Minimum Hu Diameter Density
Numbe Group Group Group Group Group Group Group Group Group Group Group Group
r I II I II I II I II I II I II
1 282 228 1002.5 1096 1717 1786 182 30 25 22 40.1 50.3
2 87 72 895.1 989 1649 1649 124 69 15 13 59.67 73.4
3 332 135 1102.9 1305 1651 1681 148 25 27 17 40.8 75.52
4 85 67 975.7 1249 1859 1646 85 297 16 14 60.98 87.9
5 94 63 1055.9 1182 1822 1834 94 40 15 14 68 87.3
6 225 192 870.5 882 1385 1421 162 18 20 29 43.5 30.6
7 145 107 712.4 993 1295 1388 150 49 19 16 37.5 60.8
8 83 54 1017 1114 1522 1579 168 89 13 15 78.2 73.21
9 46 51 744.8 673 1570 1570 130 7 10 11 74.48 61.1
10 103 88 1054.3 1146 1728 1728 147 35 13 17 81.3 64.5
11 71 71 950.83 1039 1586 1586 156 16 13 15 73.1 68.5
1
12 43 40 750.06 648 1370 1268 192 16 10 10 75 63.8
13 1000 711 1301.5 1387 1855 1863 208 43 41 47 31.7 29.1
14 73 73 933.4 972 1548 1535 167 31 16 19 58.3 50.6
15 112 112 724.8 713 1486 1513 130 113 19 15 38.1 46.22
16 240 296 847.1 1178 1495 1613 166 5 44 40 19.2 29.4
17 127 110 988.5 996 1743 1738 104 2 15 18 65.9 55.3
BJU Int., vol. 113, no. 4, pp. 610–614, 2014, doi:
V. CONCLUSIONS 10.1111/bju.12456.
[4] A. Gallioli et al., “Clinical utility of computed tomography
The proposed study was developed to provide Hounsfield characterization for percutaneous
nephrolithotomy: A cross-sectional study,” BMC Urol., vol.
technical support in kidney stone detection and diagnosis 17, no. 1, pp. 1–7, 2017, doi: 10.1186/s12894-017-0296-1.
Its function to pinpoint the CT values (area, average HU, [5] G. Bandi, R. J. Meiners, P. J. Pickhardt, and S. Y. Nakada,
density HU, diameter, minimum and maximum HU) as “Stone measurement by volumetric three-dimensional
these values provide focused investigation for medical computed tomography for predicting the outcome after
extracorporeal shock wave lithotripsy,” BJU Int., vol. 103,
specialists. no. 4, pp. 524–528, 2009, doi: 10.1111/j.1464-
The program can choose the largest slice and calculate 410X.2008.08069.x.
the accurate CT values on the selected slice. Compare [6] D.-Y. Kim and J.-W. Park, “Computer-Aided Detection of
Kidney Tumor on Abdominal Computed Tomography
with manual record, it is more accurate because the Scans,” Acta radiol., vol. 45, no. 7, pp. 791–795, 2004, doi:
whole process was developed by automatically. 10.1080/02841850410001312.
Moreover, the manual process is taken long time to select [7] R. Srisha and a M. Khan, “Morphological Operations for
the largest slice and to calculate the CT values. The Image Processing : Understanding and its Applications
Morphological Operations for Image Processing :
proposed program can reduce the processing time of Understanding and its Applications,” NCVSComs-13, no.
manual record measuring. February 2015, pp. 17–19, 2013.
Moreover, the same results are always produced for [8] L. Shapiro and G. Stockman, “Computer Vision, 1st ed.,” 1st
anyone calculation. It can measure in any circular shape ed., no. ch. 3, 2000, pp. 63–97.
[9] R. C. Gonzalez and R. E. Woods, “Digital Image Processing,
and irregular shape since the program is not depend on 3rd ed.,.” p. ch. 2, 102-107, 1992.
the shape.
Therefore, the program’s ability to select the largest
slice and to identify stone CT values provide more
efficient analysis for kidney stone diagnosis. Therefore,
the proposed is more useful compare than manual
recording since it can overcome the trade-off between the
time take and the accuracy of the manual recording.
References
[1] K. Viswanath and R. Gunasundari, “Design and analysis
performance of kidney stone detection from ultrasound image
by level set segmentation and ANN classification. In
Advances in Computing, Communications and Informatics,”
ICACCI Int. Conf. Adv. Comput. Informatics, pp. 407–414,
2014.
[2] W. Brisbane, M. R. Bailey, and M. D. Sorensen, “An
overview of kidney stone imaging techniques,” Nat. Rev.
Urol., vol. 13, no. 11, pp. 654–662, 2016, doi:
10.1038/nrurol.2016.154.
[3] W. Finch, R. Johnston, N. Shaida, A. Winterbottom, and O.
Wiseman, “Measuring stone volume - Three-dimensional
software reconstruction or an ellipsoid algebra formula?,”

You might also like