Analysis of the Use of
K-Means Clustering Method in Brain Tumor MRI Segmentation
Hanifah Fitri Maharani1*,
Lina Choridah2, Darmini3, Fatimah4, Yeti
Kartikasari5, Gatot Murti Wibowo6
1,3,4,5,6 Politeknik Kesehatan
Kementerian Kesehatan Semarang, Indonesia
2 Universitas
Gadjah Mada, Indonesia
Email: hanifahf58@gmail.com,
linachoridah@ugm.ac.id, da12mini@gmail.com, fatimah_yunaeza@yahoo.com,
yeti.kartikasari@gmail.com, gatotmurtiw@gmail.com
|
KEYWORDS |
ABSTRACT |
|
K-Means
Clustering, Linear Measurement, MRI Brain, Tumor. |
Accurate
measurement of brain tumor volume in MRI examinations is critical for
diagnosis and treatment planning. Traditionally, linear measurement is the
gold standard, but it is prone to errors due to subjectivity and fatigue, and
only provides a rough estimate of tumor volume. This research aims to compare
brain tumor volume calculation on MRI images using linear measurement and
k-means clustering method on post-contrast T1WI sequences to evaluate their
accuracy and clinical consistency. Using a quasi-experimental design with
post-test only and no control group, 32 MRI images of brain tumors were
analyzed. Tumor volumes were calculated using both methods, and the results
were statistically compared. The average tumor volume was 39,304.55 mm³ for
the linear method and 35,374.69 mm³ for the k-means clustering method.
Statistical analysis using the Wilcoxon test showed no significant difference
between the two methods (p = 0.082; p > 0.05). The results of this research
suggest that although both methods produce comparable volume estimates,
k-means clustering offers the advantage of reducing subjectivity, indicating
its potential to improve consistency and reliability of measurements in
clinical practice. The implication of this research is that the k-means
clustering method may be a more reliable alternative in the measurement of
brain tumor volume on MRI examinations, especially in reducing the
subjectivity bias often present in linear measurement methods. This can help
improve the accuracy and consistency of measurement results, which is crucial
for more precise treatment planning and evaluation of patient therapy
response. |
|
DOI: 10.58860/ijsh.v3i11.264 |
|
Corresponding Author: Hanifah Fitri Maharani*
Email: hanifahf58@gmail.com
INTRODUCTION
A tumor is an
abnormal growth of intracranial tissue or the meninges, which can be classified
into benign and malignant types based on the degree of malignancy
Diagnosis of
brain tumors can be achieved through several non-invasive imaging modalities,
such as MRI, PET, and SPECT. For MRI scans, gadolinium contrast media is
typically administered to enhance image clarity and resolution. Post-contrast
MRI images are captured in multiple sequences, including axial T1, sagittal T1,
and coronal T1 views
The
measurement results using linear measurement are considered a rough and
unspecific assessment that cannot show the volume of the tumor. Therefore, an
image segmentation system has been developed in the development of medical
imaging technology. Image segmentation has an important role in the extraction,
analysis, and interpretation of an image that is widely applied in medicine.
For example, tissue classification, tumor localization, estimating tumor
volume, surgical planning, depiction of blood cells, and image
registration
Image
segmentation can be applied to 2D and 3D sequences, there are several image
segmentation methods, one of which is the K-means clustering method. This
method is one of the faster and more sensitive methods than other methods, such
as the FCM method. In addition, this method is also a simple method whose
working principle is to divide image pixels into several cluster groups
according to their characteristics and is able to overcome the fuzziness
that comes from grayscale images
This research
introduces a post-processing approach in MRI image analysis, specifically
employing the K-means clustering segmentation method to measure brain tumor
volume
Conducting
tumor volume calculations on MRI brain tumor examinations using both the linear
measurement method and the K-means clustering method, with the goal of
comparing these approaches
Based on the
above background, this research aims to determine the advantages in efficiency
and accuracy offered by K-means clustering over traditional linear measurement,
which has the potential to improve MRI reading time and follow-up assessment
accuracy for clinical radiology practice. Thus, the benefit of this research is
to contribute to the development of a more accurate and consistent method of
measuring brain tumor volume on MRI examinations, particularly to improve the
work efficiency of radiology specialists and the quality of clinical
assessment. The K-means clustering method can be a more reliable alternative to
reduce subjectivity and human error, thereby increasing confidence in the
results of brain tumor volume measurements.
METHOD
This quasi-experimental research employs a
post-test-only design without a control group, analyzing a sample of 32 MRI
images from brain tumor patients. Brain tumor volume was measured using K-means
clustering segmentation on T1-weighted post-contrast spin echo sequences in
axial cuts. Linear measurements were conducted on T1-weighted post-contrast
spin echo sequences in both axial and sagittal cuts. For reproducibility,
Matlab software was used for the K-means clustering, and the Siemens 1.5T MRI
modality was chosen to optimize imaging quality.
RESULT AND DISCUSSION
Analysis of Tumor Volume
Calculation Using Linear Measurement Method on 1.5T MRI
The research
on brain tumor volume calculation using the linear measurement method on MRI
was conducted twice by radiology specialists at separate intervals, employing a
quantitative approach. This calculation method uses two perpendicular
cross-sectional slices, specifically the axial and sagittal views. The selected
slice is the one with the largest tumor dimension. The process for calculating
brain tumor volume using the linear measurement method includes the following
steps:
1.
The
workstation display is configured to a 2x1 layout to facilitate comparative
imaging.
2.
Axial slices
are chosen by selecting the T1WI post-contrast axial slice sequence (t1_se_tra
CE) and displayed on the left layout, while the right layout is reserved for
displaying the T1WI post-contrast sagittal sequence (t1_se_sag CE).
3. Selection
of the slice that will be measured to obtain the largest tumor dimension in
both cross-sectional cuts. Slice selection is done by scrolling on both images
used.
4. Measurement
with linear measurement is done through the distance tool available on the view
toolbar menu. In axial cut images, tumor diameter measurement is performed by
drawing a vertical line from the front edge to the back edge of the tumor area,
and then Dap (anterior-posterior diameter) will be obtained. Next, a
measurement is made with a horizontal line in the tumor area; then, the Dl
(lateral diameter) size will be obtained. An example of measuring the diameter
of a brain tumor with linear measurement can be seen in Figure 1.
Figure 1. Example of Diameter
Measurement Results in a Brain Tumor
Using Linear Measurement
Method
5.
In the image
with a sagittal cut, measurements are made by making a vertical line from the
top to the base of the tumor; from these measurements will get Dcc
(craniocaudal diameter).
6.
After all the
diameter sizes have been obtained, then proceed with the calculation of the
brain tumor volume using the ellipsoid formula, namely:
V = DCC x dl x dap x 
Description:
DCC : craniocaudal diameter
Dap : anteroposterior diameter
Dl :
lateral diameter.
7.
The
procedure is carried out on all samples used so that it will get the results of
brain tumor volume in 32 image samples used
Analysis of Tumor Volume Calculation Using K-Means Clustering
Method
The research results on the calculation of brain tumor volume using the
k-means clustering segmentation method, starting with the creation of machine
learning k-means clustering in the Matlab program. Calculation of brain tumor
volume with the k-means clustering method is done with several stages, namely:
1.
Select
patient DICOM data by clicking the browse folder tool. The image used is an
axial image with the t1_se_tra CE sequence. The image used is a slice that has
an image of a tumor in the brain.
2.
The
selected image is displayed in box 1 as the original image, then click process
to run the image segmentation program. The segmented image will be displayed in
box 2 in the form of a ground truth image, while the 3D visualization of the
image will be displayed in box 3, and the tumor volume will appear in Vol. The
following is the initial appearance of the K-Means Clustering segmentation
program in Figure 2.
Figure 2. Initial View
of the K-Means Clustering Segmentation Program
Table 1. Window Functions of the Segmentation Program
|
Window |
Function |
|
Window 1 |
Display the original MRI
image |
|
Window 2 |
Display image of brain tumor segmentation result |
|
Window 3 |
Display the brain tumor
segmentation result in 3D |
|
Button functions in the segmentation program |
|
|
Window |
Function |
|
Browser |
Select the MRI image folder |
|
Process |
To start brain tumor segmentation |
|
Vol |
Display brain tumor volume
results |
|
Save |
To store the segmented image and
volume |
|
Reset |
To clear the working
window for evaluation on another image |
Examples of image segmentation results with the
k-means clustering method displaying the original image, segmented image, and
3D visualization of brain tumors. This can be seen in Figure 3.
Figure 3. Examples of the original image
(a), segmented image (b), ground truth image
(c),
and 3D visualization of the tumor.
The
documentation of the brain tumor volume calculation with the K-means clustering
method can be seen in Figure 4.
Figure 4. Example of brain tumor volume
measurement results with the method k-means clustering
Comparison of Brain Tumor Volume Calculation with Linear
Measurement and K-Means Clustering Method
Based on the results of the research that has been obtained regarding the
calculation of brain tumor volume with linear measurement and k-means
clustering methods, statistical analysis will then be carried out to determine
whether or not there is a difference in the results of calculating brain tumor
volume in the two methods. A comparison of brain tumor volume measurement
results with linear measurement and k-means clustering methods can be seen in
Table 1.
Table 2. Comparison of Brain Tumor Volume by Linear Measurement
Method and K-means Clustering
|
Sample |
Linear measurement (mm )3 |
K-means clustering (mm )3 |
|
Sample 1 |
8008.00 |
7146.61 |
|
Sample 2 |
1464.32 |
2698.24 |
|
Sample 3 |
48971.89 |
44096.79 |
|
Sample 4 |
21508.76 |
25057.01 |
|
Sample 5 |
22959.30 |
1678.96 |
|
Sample 6 |
60950.92 |
47928.96 |
|
Sample 7 |
74020.13 |
31694.46 |
|
Sample 8 |
19777.87 |
17322.3 |
|
Sample 9 |
1856.40 |
1509.45 |
|
Sample 10 |
30850.56 |
14315.19 |
|
Sample 11 |
105963.00 |
127709.47 |
|
Sample 12 |
10284.46 |
10486.45 |
|
Sample 13 |
37619.40 |
34800.99 |
|
Sample 14 |
3871.39 |
2696.01 |
|
Sample 15 |
26208.00 |
29243.48 |
|
Sample 16 |
144144.00 |
139351.68 |
|
Sample 17 |
5304.87 |
7210.22 |
|
Sample 18 |
7960.68 |
2292.85 |
|
Sample 19 |
66664.00 |
75991.33 |
|
Sample 20 |
30476.16 |
23259.18 |
|
Sample 21 |
131459.33 |
105737.85 |
|
Sample 22 |
87207.99 |
76018.8 |
|
Sample 23 |
38012.46 |
41542.6 |
|
Sample 24 |
19068.40 |
17322.3 |
|
Sample 25 |
2365.54 |
2279.66 |
|
Sample 26 |
3556.80 |
1509.45 |
|
Sample 27 |
11982.48 |
18344.97 |
|
Sample 28 |
25828.29 |
22963.62 |
|
Sample 29 |
43056.00 |
45995.36 |
|
Sample 30 |
80886.00 |
70909.62 |
|
Sample 31 |
18230.11 |
30574.95 |
|
Sample 32 |
67228.20 |
52301.51 |
Calculation of brain tumor
volume by linear measurement method obtained the largest result in sample 16
with a volume of 144144.00 mm3, while the smallest value in sample 2 with
a volume of 1464.32 mm3. Calculation of brain tumor volume with this method
obtained an average value of 39304.53 mm3. For the calculation of brain tumor
volume with the k-means clustering method, the largest volume was obtained in
sample 16 with a volume of 139351.68 mm3 and the smallest volume in sample
26 with a volume of 1509.45 mm3. The average value in the k-means clustering
method is 35374.69 mm.3
This research shows the analysis of differences
in the calculation of brain tumor volume performed using manual methods, namely
linear measurement and segmentation methods, namely k-means clustering
Data were collected from patients with clinical
brain tumors who performed MRI examinations using routine brain MRI examination
sequences. Routine brain MRI examinations in clinical brain tumors use spin
echo sequences T1 weighting sagittal cuts, spin echo sequences T1 weighting
axial cuts, spin echo sequences T2 weighting axial cuts, TIRM sequences T2
weighting axial cuts, blade sequences T2 weighting coronal cuts, spin echo
sequences T1 weighting coronal cuts. After that, contrast media was injected
into the patient, and image acquisition was performed with spin echo sequences
of T1 weighting of sagittal section + contrast media, spin echo sequences of T1
weighting of axial section + contrast media, and spin echo sequences of T1
weighting of coronal section + contrast media.
For the measurement of brain tumor volume, the
linear measurement method used T1 post-contrast weighted spin echo sequence
images on axial and sagittal sections with t1_se_tra CE and t1_se_sag CE
sequences. The use of contrast media has the benefit of clarifying the image of
the brain tumor in the image used for tumor volume calculation. In addition, in
image segmentation, the T1WI post-contrast sequence is the most commonly used
sequence because it is able to visualize the lesion well and does not show edema
areas. Thus, the resulting image and segmentation focus on the tumor area
The results of this research are quantitative
data in the form of brain tumor volume calculations that have been carried out
by radiology specialists who act as observers. An observer performed the volume
calculation method with linear measurement on 1.5 Tesla MRI modality. The
calculation of brain tumor volume with this method was performed on T1WI post-contrast
weighted spin echo sequences in axial and sagittal sections. The use of T1WI
post-contrast sequences is because these sequences can make the tumor image
clearer due to the use of contrast media given to patients through intravenous
injections. The contrast media used was Gd-DTPA.
Calculation of brain tumor volume with linear
measurement method is done by several steps, namely opening the MRI image file
of the patient's brain tumor and selecting the SE T1WI post-contrast axial
slice sequence, then the radiologist will select one slice that displays the
largest tumor area, and continue by measuring the diameter of the tumor. The
diameter measurement is done using the distance tools menu located on the view
toolbar. Measurements on the axial slice will obtain the lateral diameter (Dl) and
anterior-posterior diameter (Dap). Tumor diameter measurements were also taken
on the t1_se_sag CE sagittal slice sequence by selecting one sagittal slice
that displays the largest tumor area. The tumor diameter was then measured to
obtain the cranio-caudal diameter (Dcc). The final results of these
measurements are Dap, Dcc, and Dl. To calculate the tumor volume, the ellipsoid
formula is used, namely V = Dap x Dcc x Dl x π/6.
The stages that have been carried out to measure
the volume of brain tumors using the linear measurement method are in
accordance with the research conducted the research explained that the
calculation of tumor volume could be done by measuring the diameter of the tumor
using linear measurement and continued with the calculation of tumor volume
with the ellipsoid formula where V = Dcc x Dl x Dap x π/6 or can also be
called diameter-based volume
Calculation of brain tumor volume with k-means
clustering segmentation method using post-contrast MRI images; this is because
the algorithm used for tumor detection performs analysis based on the contrast
in MRI images. The post-contrast MRI image shows a hyperintense image of the
tumor compared to the normal area
In this research, the segmentation method and the
K-Means clustering segmentation stages are in accordance with research
conducted by Sabaghian S (2020), in which it is explained that the stages in
image segmentation begin with preparing the automatic detection module to be
used and also the image to be tested. Furthermore, anisotropic filters were provided
to the image with the aim of removing noise in the image and performing the
skull stripping process. Then, proceed with the application of the k-means clustering
algorithm on the image and determine the number of clusters to be used. In this
research, the number of clusters used is 3 clusters, so K = 3. The MRI image
section is divided into 3 clusters, namely the tumor area, the normal
area, and also the background area
Calculation of brain tumor volume using linear
measurement method and k-means clustering method yielded different volume
results. The comparison of the volume calculation results was obtained because
of the different ways of calculating the volume used. In the linear measurement
method, the tumor volume calculation is done by measuring the diameter of one
axial slice image that has the largest tumor dimensions to obtain the length
(p) x width (l) of the tumor, while the height (t) of the tumor is measured
through one of the sagittal slice images that have the largest tumor
dimensions. From the measurements on the two perpendicular slices, the size of
p x l x t or Dap x Dcc x Dl of the tumor is obtained, and then the tumor volume
calculation will be carried out with the ellipsoid formula. Furthermore, the
calculation of tumor volume with the k-means clustering segmentation method is
carried out on axial slice images on all slices that have tumor images. This is
in accordance with the research, the k-means clustering segmentation method,
the calculation of tumor volume is obtained from the thickness of the slice
thickness used, namely using a 5 mm slice thickness in MRI examination of brain
tumors. Thus, to obtain brain volume results, each tumor area is multiplied by
the slice thickness used and then accumulated.
The Wilcoxon test showed that there was no
significant difference in the calculation of brain tumor volume with linear
measurement and k-means clustering methods with a p-value > 0.05, namely
0.082. The results of the calculation of brain tumor volume with the two
methods are not significantly different, although the measurement principles of
the two methods used are different. Measurement of brain tumor volume with
linear measurement is only performed on one slice that visualizes the largest tumor
area on two perpendicular cross-sectional pieces. Measurements are taken to
obtain the length, width, and height of the tumor or Dcc x Dl x Dap. The results of these measurements are
multiplied by a constant 0.52, according to the ellipsoids formula, which is
Dcc x Dl x Dap x 0.52. This step is in accordance with the research of
regarding the calculation of tumor volume using linear measurement
In calculating the volume of
brain tumors with both methods, there are several samples with relatively small
differences in brain tumor volume results or <5000mm3. One example is in
sample 1, where the volume result with linear measurement is 8008.00mm3, while
with k-means clustering is 7146.61mm3, the difference in tumor volume
calculation results is 861.39mm3. The MRI image of sample 1 can be seen in
Figure 5. The image of sample 1 shows a tumor that is regular or close to
a round shape and has almost the same shape in each slice. So, the results
of calculating tumor volume with the ellipsoid formula and the results of
calculating tumor volume with k-means clustering tend to be the same.
Figure 5. MRI Image of Sample 1 Brain Tumor Showing Regular Tumor Shape
The results of this research
show that the calculation of brain tumor volume with the linear measurement
method is generally greater than the k-means clustering method. The comparison
of brain tumor volume calculations using linear measurement and segmentation
methods showed that the results of tumor volume calculations with linear
measurement were greater than the results of volume calculations with image
segmentation. This is due to the difference in the size of t or lateral
diameter (Dl), in the calculation with linear measurement, the value of t is
obtained from the largest diameter of the sagittal slice, while in the image
segmentation method, the value of t is the result of the number of slices that
have tumor images with the slice thickness used
One example of a brain tumor
volume result with a large difference is in sample 21, where the volume result
with linear measurement is 131459.33mm3, while with k-means clustering is
105737.85mm3 with a difference of up to 25721.48 mm3. This is because the
method of measuring the volume of brain tumors with linear measurement uses the
ellipsoid formula, where the object measured or the tumor measured is the tumor
with the largest area that is considered to represent the shape of the tumor as
a whole and is assumed to be in the form of an ellipse that can be calculated
volume. The k-means clustering method does not measure the volume of the object
based on the shape of the space but calculates the tumor area in each slice
based on the image pixel. Therefore, the shape of the segmentation results
follows the intensity of the image or the shape of the tumor based on the
enhancement of the image. So, the
difference in the results of measuring the volume of brain tumors in the two
methods can be explained by the difference in volume calculation principles in
the two methods used, namely linear measurement techniques that refer to the
shape of an elliptical space and k-means clustering which does not refer to a
specific space shape but by measuring the segmentation area of each object and
visualized in 3D
Figure 6. MRI Image of Brain Tumor Sample 21 Showing Irregular Tumor Shape
In 11 samples used, the results
of brain tumor volume with the K-means clustering method are greater than the
results of tumor volume with linear measurement. For example, in sample 11, the
volume result with the k-means clustering method is 127709.47mm3, and the
linear measurement method is 105963.00mm3. The difference between the two
methods reached 21746.47mm.3 The
magnitude of the results of measuring the volume of brain tumors with k-means
clustering can be caused by tumors that have spread to the brain area; this is
the nature of neoplasms from tumors, so the segmentation area needed is wider.
Segmentation of brain tumor MRI images with
K-means clustering has many benefits in tumor segmentation. In addition, the
use of K-means clustering can still be developed to get more accurate
segmentation results, such as the use of K-means clustering combined with
Convolutional Neural Network (CNN) or deep learning for more specific brain tumor
detection to classify the level of tumor malignancy (benign or malignant tumors).
K-means clustering can also be developed using the iterative Co-Clustering and
K-Means (ICCK) algorithm. The working principle of Co-Clustering is clustering
the rows and columns of the matrix. This ICCK segmentation method obtained
better results on more complex images and also has high sensitivity,
specificity, and accuracy values of 82.41%, 99.74%, and 99.28% in brain tumor
MRI image detection
The advantage of the linear measurement method in
the MRI examination of brain tumors is that it uses linear measurement as a
gold standard to determine the size of the tumor performed by radiology
specialists
The advantage of the k-means clustering
segmentation method is unsupervised learning, which has a function to divide
the pixel intensity in the image; besides that, this method is an automatic
segmentation method
Digital image processing is an important part of
efforts to enforce diagnoses on patients, such as in measuring the volume of
brain tumors owned by patients. In this research, the digital image processing
method used, especially in the image segmentation method, produces an effective
tumor volume calculation that can help radiology specialists, although the
application of this segmentation method is still limited to images with post-contrast
T1WI weighting in the spin echo sequence. The resulting image segmentation is
able to reduce subjectivity because the segmentation is done based on pixel
intensity. In addition, the use of k-means clustering segmentation is able to
improve patient diagnosis information and can be an alternative in calculating
brain tumor volume
In this research, there are several limitations
in the application of the k-means clustering segmentation method, which is only
applied to one modality of MRI aircraft, namely Siemens 1.5T MRI
CONCLUSION
In analyzing
the application of the K-means clustering method in MRI segmentation of brain
tumors, a comparison was made between the linear measurement method and K-means
clustering for calculating tumor volume. The linear measurement method
estimates the volume using the largest single slice from the axial and sagittal
views, while K-means clustering considers the entire set of slices. Although
these methods yielded different volume estimates, the Wilcoxon statistical test
indicated that the difference was not statistically significant. The linear
measurement method generally provided slightly higher volume estimates than the
K-means clustering method, but this difference lacked statistical significance.
For further development, it is recommended to optimize the K-means clustering
method, especially in the segmentation of brain tumors with irregular edges, in
order to more accurately distinguish tumors from surrounding normal tissues.
REFERENCES
Biratu, E. S., Schwenker, F., Ayano, Y. M., & Debelee, T. G. (2021).
A survey of brain tumor segmentation and classification algorithms. Journal
of Imaging, 7(9), 179.
Chaulagain, D., Smolanka, V., & Smolanka, A. (2021). Intracranial
tumors: overview, histological types, symptoms and treatment plans. Int. J.
Health Sci. Res. 2021. 11 (10). 133, 144.
Choi, Y. K., Lee, J. H., Kim, Y. S., Wee, C. W., Kim, Y. B., Park, S.
J., Jung, W., Seol, K. H., & Choi, E. (2024). Mid-treatment MRI-based
tumor response assessment for tumor recurrence and patient survival in locally
advanced adenocarcinoma of the cervix: A retrospective
multicenter research of KROG 23-03. Gynecologic Oncology, 190,
222–229. https://doi.org/10.1016/j.ygyno.2024.08.025
Chouhan, S. S., Kaul, A., & Singh, U. P. (2019). Image segmentation
using computational intelligence techniques. Archives of Computational
Methods in Engineering, 26, 533–596.
Dhanachandra, N., Manglem, K., & Chanu, Y. J. (2015). Image
Segmentation Using K -means Clustering Algorithm and Subtractive Clustering
Algorithm. Procedia Computer Science, 54,
764–771. https://doi.org/10.1016/j.procs.2015.06.090
Jindal, S. K., Banerjee, S., Patra, R., & Paul,
A. (2022). Deep
learning-based brain malignant neoplasm classification using MRI image
segmentation assisted by bias field correction and histogram equalization. In Brain
Tumor MRI Image Segmentation Using Deep Learning Techniques (pp. 135–161).
Elsevier. https://doi.org/10.1016/B978-0-323-91171-9.00008-9
Kaya, I. E., Pehlivanlı, A. Ç., Sekizkardeş, E. G., &
Ibrikci, T. (2017). PCA based clustering for brain tumor segmentation of T1w
MRI images. Computer Methods and Programs in Biomedicine, 140,
19–28. https://doi.org/10.1016/j.cmpb.2016.11.011
Khalil, Y. Al, Ayaz, A., Lorenz, C., Weese, J., Pluim, J., &
Breeuwer, M. (2024). Multi-modal brain tumor segmentation via conditional
synthesis with Fourier domain adaptation. Computerized Medical Imaging and
Graphics, 112, 102332.
https://doi.org/10.1016/j.compmedimag.2024.102332
Li, J., Erdt, M., Janoos, F., Chang, T., & Egger, J. (2021). Medical
image segmentation in oral-maxillofacial surgery. In Computer-Aided Oral
and Maxillofacial Surgery (pp. 1–27). Elsevier.
https://doi.org/10.1016/B978-0-12-823299-6.00001-8
Mohammadi, S., Ghaderi, S., Ghaderi, K., Mohammadi, M., & Pourasl,
M. H. (2023). Automated segmentation of meningioma from contrast-enhanced
T1-weighted MRI images in a case series using a marker-controlled watershed
segmentation and fuzzy C-means clustering machine learning algorithm. International
Journal of Surgery Case Reports, 111, 108818.
https://doi.org/10.1016/j.ijscr.2023.108818
Nasor, M., & Obaid, W. (2020). Detection and Localization of
Early‐Stage Multiple Brain Tumors Using a Hybrid Technique of
Patch‐Based Processing, k‐means Clustering and Object Counting. International
Journal of Biomedical Imaging, 2020(1), 9035096.
Perkins, A., & Liu, G. (2016). Primary brain tumors in adults:
diagnosis and treatment. American Family Physician, 93(3),
211-217B.
Putri, M. N., Adi, K., Katili, M. I., Hariyanto, S., & Rochmayanti,
D. (2022). Comparison Measurement And Calculation
Of Brain Tumor In Mri Modalities Utilising Spin-Echo Pulse Sequence
T1-Weighted Contrast And Digital Image Processing Applications.
Sabaghian, S., Dehghani, H., Batouli, S. A. H., Khatibi, A., &
Oghabian, M. A. (2020). Fully automatic 3D segmentation of the thoracolumbar
spinal cord and the vertebral canal from T2-weighted MRI using K-means
clustering algorithm. Spinal Cord, 58(7),
811–820.
Sarris, A. L., Sidiropoulos, E., Paraskevopoulos,
E., & Bamidis, P. (2023). Towards a Digital Twin in Human Brain: Brain Tumor Detection
Using K-Means. In Caring is Sharing–Exploiting the Value in Data for Health
and Innovation (pp. 1052–1056). IOS Press.
Sreenivasan, S. A., Madhugiri, V. S., Sasidharan, G. M., & Kumar, R.
V. R. (2016). Measuring glioma volumes: a comparison of linear measurement based formulae with the manual image
segmentation technique. Journal of Cancer Research and Therapeutics, 12(1),
161–168.
Vishnuvarthanan, G., Rajasekaran, M. P., Subbaraj, P., &
Vishnuvarthanan, A. (2016a). An unsupervised learning method with a clustering
approach for tumor identification and tissue segmentation in magnetic
resonance brain images. Applied Soft Computing, 38, 190–212.
https://doi.org/10.1016/j.asoc.2015.09.016
Vishnuvarthanan, G., Rajasekaran, M. P., Subbaraj, P., &
Vishnuvarthanan, A. (2016b). An unsupervised learning method with a clustering
approach for tumor identification and tissue segmentation in magnetic
resonance brain images. Applied Soft Computing, 38, 190–212.
https://doi.org/10.1016/j.asoc.2015.09.016
Wadhwa, A., Bhardwaj, A., & Singh Verma, V. (2019a). A review on
brain tumor segmentation of MRI images. Magnetic Resonance Imaging, 61,
247–259. https://doi.org/10.1016/j.mri.2019.05.043
Wadhwa, A., Bhardwaj, A., & Singh Verma, V. (2019b). A review on
brain tumor segmentation of MRI images. Magnetic Resonance Imaging, 61,
247–259. https://doi.org/10.1016/j.mri.2019.05.043
|
|
|
