• International Journal of Computer Science & Network Security
• /
• v.22 no.8
• /
• pp.343-351
• /
• 2022

A brain tumor forms when some tissue becomes old or damaged but does not die when it must, preventing new tissue from being born. Manually finding such masses in the brain by analyzing MRI images is challenging and time-consuming for experts. In this study, our main objective is to detect the brain's tumorous part, allowing rapid diagnosis to treat the primary disease instantly. With image processing techniques and deep learning prediction algorithms, our research makes a system capable of finding a tumor in MRI images of a brain automatically and accurately. Our tumor segmentation adopts the U-Net deep learning segmentation on the standard MICCAI BRATS 2018 dataset, which has MRI images with different modalities. The proposed approach was evaluated and achieved Dice Coefficients of 0.9795, 0.9855, 0.9793, and 0.9950 across several test datasets. These results show that the proposed system achieves excellent segmentation of tumors in MRIs using deep learning techniques such as the U-Net algorithm.

• The Journal of Korean Society for Radiation Therapy
• /
• v.26 no.1
• /
• pp.127-135
• /
• 2014

Purpose : By taking advantage of each imaging modality, the use of fused CT/MRI image has increased in prostate cancer radiation therapy. However, fusion uncertainty may cause partial target miss or normal organ overdose. In order to complement such limitation, our hospital acquired MRI image (Planning MRI) by setting up patients with the same fixing tool and posture as CT simulation. This study aims to evaluate the usefulness of the Planning MRI through comparing and analyzing the diagnostic MRI image and Planning MRI image. Materials and Methods : This study targeted 10 patients who had been diagnosed with prostate cancer and prescribed nonhormone and definitive RT 70 Gy/28 fx from August 2011 to July 2013. Each patient had both CT and MRI simulations. The MRI images were acquired within one half hour after the CT simulation. The acquired CT/MRI images were fused primarily based on bony structure matching. This study measured the volume of prostate in the images of Planning MRI and diagnostic MRI. The diameters at the craniocaudal, anteroposterior and left-to-right directions from the center of prostate were measured in order to compare changes in the shape of prostate. Results : As a result of comparing the volume of prostate in the images of Planning MRI and diagnostic MRI, they were found to be $25.01cm^3$(range $15.84-34.75cm^3$) and $25.05cm^3$(range $15.28-35.88cm^3$) on average respectively. The diagnostic MRI had an increase of 0.12 % as compared with the Planning MRI. On the planning MRI, there was an increase in the volume by $7.46cm^3$(29 %) at the transition zone directions, and there was a decrease in the volume by $8.52cm^3$(34 %) in the peripheral zone direction. As a result of measuring the diameters at the craniocaudal, anteroposterior and left-to-right directions in the prostate, the Planning MRI was found to have on average 3.82cm, 2.38cm and 4.59cm respectively and the diagnostic MRI was found to have on average 3.37cm, 2.76cm and 4.51cm respectively. All three prostate diameters changed and the change was significant in the Planning MRI. On average, the anteroposterior prostate diameter decrease by 0.38cm(13 %). The mean right-to-left and craniocaudal diameter increased by 0.08cm(1.6 %) and 0.45cm(13 %), respectively. Conclusion : Based on the results of this study, it was found that the total volumes of prostate in the Planning MRI and the diagnostic MRI were not significantly different. However, there was a change in the shape and partial volume of prostate due to the insertion of prostate balloon tube to the rectum. Thus, if the Planning MRI images were used when conducting the fusion of CT/MRI images, it would be possible to include the target in the CTV without a loss as much as the increased volume in the transition zone. Also, it would be possible to reduce the radiation dose delivered to the rectum through separating more clearly the reduction of peripheral zone volume. Therefore, the author of this study believes that acquisition of Planning MRI image should be made to ensure target delineation and localization accuracy.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.17 no.2
• /
• pp.10-14
• /
• 2013

Purpose: Due to developed simultaneous PET/MRI, it has become possible to obtain more anatomical image information better than conventional PET/CT. By the way, in the PET/CT, the linear absorption coefficient is measured by X-ray directly. However in case of PET/MRI, the value is not measured from MRI images directly, but is calculated by dividing as 4 segmentation ${\mu}-map$. Therefore, in this paper, we will evaluate the SUV's difference of attenuation correction PET images from PET/MRI and PET/CT. Materials and Methods: Biograph mCT40 (Siemens, Germany), Biograph mMR were used as a PET/CT, PET/MRI scanner. For a phantom study, we used a solid type $^{68}Ge$ source, and a liquid type $^{18}F$ uniformity phantom. By using VIBE-DIXON sequence of PET/MRI, human anatomical structure was divided into air-lung-fat-soft tissue for attenuation correction coefficient. In case of PET/CT, the hounsfield unit of CT was used. By setting the ROI at five places of each PET phantom images that is corrected attenuation, the maximum SUV was measured, evaluated %diff about PET/CT vs. PET/MRI. In clinical study, the 18 patients who underwent simultaneous PET/CT and PET/MRI was selected and set the ROI at background, lung, liver, brain, muscle, fat, bone from the each attenuation correction PET images, and then evaluated, compared by measuring the maximum SUV. Results: For solid $^{68}Ge$ source, SUV from PET/MRI is measured lower 88.55% compared to PET/CT. In case of liquid $^{18}F$ uniform phantom, SUV of PET/MRI as compared to PET/CT is measured low 70.17%. If the clinical study, the background SUV of PET/MRI is same with PET/CT's and the one of lung was higher 2.51%. However, it is measured lower about 32.50, 40.35, 23.92, 13.92, 5.00% at liver, brain, muscle, fat, femoral head. Conclusion: In the case of a CT image, because there is a linear relationship between 511 keV ${\gamma}-ray$ and linear absorption coefficient of X-ray, it is possible to correct directly the attenuation of 511 keV ${\gamma}-ray$ by creating a ${\mu}$map from the CT image. However, in the case of the MRI, because the MRI signal has no relationship at all with linear absorption coefficient of ${\gamma}-ray$, the anatomical structure of the human body is divided into four segmentations to correct the attenuation of ${\gamma}-rays$. Even a number of protons in a bone is too low to make MRI signal and to localize segmentation of ${\mu}-map$. Therefore, to develope a proper sequence for measuring more accurate attenuation coefficient is indeed necessary in the future PET/MRI.

• Journal of the Korean Society of Radiology
• /
• v.17 no.7
• /
• pp.1123-1131
• /
• 2023

This study aimed to present a method to effectively suppress metal artifacts caused by spinal fusion surgery during spinal MRI study. For this purpose, a phantom made of spinal surgery screws was created to reproduce the metal artifact. Then, images were acquired with 1.5T and 3.0T MRI to evaluate changes in metal artifacts according to magnetic field strength. In addition, metal artifacts were evaluated by increasing the receive bandwidth to 200, 400 and 800 Hz/PX. As a result, metal artifacts occurring in images obtained from the 1.5T MRI decreased by approximately 52.2% compared to images obtained from the 3.0T MRI, showing a significant difference (p<0.05). In particular, the signal loss and signal pile up areas were reduced by approximately 52.81% and 42.71%, respectively, showing a significant effect in suppressing metal artifacts. On the other hand, when images were acquired while increasing the receive bandwidth from 200 to 800 Hz/PX, there was no significant effect, with a decrease of up to 8.93% for the 1.5T MRI and up to 10.98% for the 3.0T MRI (p>0.05). As a result of this study, increasing the receive bandwidth reduced signal loss and reduced some metal artifacts, but did not have a significant effect because it did not suppress signal pile up. However, when the magnetic field strength was reduced from 3.0T to 1.5T, signal loss and signal pile up were greatly reduced, effectively improving the metal artifact. Therefore, in order to suppress metal artifacts caused by spinal fusion surgery, study using a low magnetic field MRI can be said to be the most effective method.

• Journal of the Korean Society of Radiology
• /
• v.1 no.2
• /
• pp.23-30
• /
• 2007

This paper presents a method of generating 3-dimensional images by preprocessing 2-dimensional abdominal images obtained using CT (computed tomography) and MRI (magnetic resonance imaging) through segmentation, threshold technique, etc. and apply the method to virtual endoscopy. Three-dimensional images were visualized using indirect volume rendering, which can render at high speed using a general-purpose graphic accelerator used in personal computers. The algorithm used in the rendering is Marching Cubes, which has only a small volume of calculation. In addition, we suggested a method of producing 3-dimensional images in VRML (virtual reality modeling language) running on the Web browser without a workstation or an exclusive program. The number of nodes, the number of triangles and the size of a 3-dimensional image file from CT were 85,367, 174,150 and 10,124, respectively, and those from MRI were 34,029, 67,824 and 3,804, respectively.

• Progress in Medical Physics
• /
• v.24 no.4
• /
• pp.237-242
• /
• 2013

In the intracranial regions, an accurate delineation of the target volume has been difficult with only the CT data due to poor soft tissue contrast of CT images. Therefore, the magnetic resonance images (MRI) for the delineation of the target volumes were widely used. To calculate dose distributions with MRI-based RTP, the electron density (ED) mapping concept from the diagnostic CT images and the pseudo CT concept from the MRI were introduced. In this study, the look up table (LUT) from the fifteen patients' diagnostic brain MRI images was created to verify the feasibility of MRI-based RTP. The dose distributions from the MRI-based calculations were compared to the original CT-based calculation. One MRI set has ED information from LUT (lMRI). Another set was generated with voxel values assigned with a homogeneous density of water (wMRI). A simple plan with a single anterior 6MV one portal was applied to the CT, lMRI, and wMRI. Depending on the patient's target geometry for the 3D conformal plan, 6MV photon beams and from two to five gantry portals were used. The differences of the dose distribution and DVH between the lMRI based and CT-based plan were smaller than the wMRI-based plan. The dose difference of wMRI vs. lMRI was measured as 91 cGy vs. 57 cGy at maximum dose, 74 cGt vs. 42 cGy at mean dose, and 94 cGy vs. 53 at minimum dose. The differences of maximum dose, minimum dose, and mean dose of the wMRI-based plan were lower than the lMRI-based plan, because the air cavity was not calculated in the wMRI-based plan. These results prove the feasibility of the lMRI-based planning for brain tumor radiation therapy.

• Progress in Medical Physics
• /
• v.21 no.4
• /
• pp.348-353
• /
• 2010

The purpose of this study was to explore the potentials of a clinical 3T MRI in mouse brains and technical adaptation and optimization. T1-weighted images (T1WI), T2-weighted images (T2WI), FLAIR (Fluid Attenuated Inversion Recovery) images, Gadolinium enhanced T1-weighted images (Gd-T1WI), Diffusion weighted images (DWI) were acquired in brain of 2 mice (weight 20~25 g) with cerebral infarction by occlusion of right middle cerebral artery, 1 hour, 24 hours, 72 hours after infarction and 1 normal mouse brain using clinical 3T MRI scanner. We analyzed differentiation of striatum, ventricle, cerebral cortex, and possibility of detection of acute cerebral infarction. We could differentiate the striatum, ventricle, cerebral cortex on T2WI and on DWI, FLAIR, T1WI, the differentiation of each anatomy of brain was not definite, but acute cerebral infarction was detected on DWI of 1 hour, 24 hours, 72 hours after infarction and on T2WI, FLAIR of 24 hours, 72 hours after infarction. Clinical 3T MRI can be used in differentiation of anatomy of mouse brains and DWI can be helpul in detection of acute cerebral infarction in acute phase. With technical adaptation and optimization clinical 3T MRI can be useful tool for provide preclinical and clinical small animal studies.

• Proceedings of the IEEK Conference
• /
• 2002.07b
• /
• pp.802-805
• /
• 2002

In the three-dimensional domain image expressed with two-dimensional slice images, such as fMRI images and multi-slice CT images, we propose the three-dimensional domain automatic segmentation for the purpose of extracting region. In this paper, we segmented each domain from the fMRI images of the head of people and monkey. We used the neural network "Pulse-Coupled Neural Network" which is one of the models of visual cortex of the brain based on the knowledge from neurophysiology as the technique. By using this technique, we can segment the region without any learning. Then, we reported the result of division of each domain and extraction to the fMRI slice images of human's head using "three-dimensional Pulse-Coupled Neural Network" which is arranged and created the neuron in the shape of a three-dimensional lattice.

• Journal of the Korean Society of Radiology
• /
• v.79 no.6
• /
• pp.340-347
• /
• 2018

Purpose: To evaluate value of image subtraction for identifying hepatocellular carcinoma (HCC) capsule on gadoxetic acid-enhanced MR images. Materials and Methods: This study involved 108 patients at risk of HCC preoperatively examined using gadoxetic acid-enhanced MRI with hepatic resection between May 2015 and February 2017. We evaluated qualities of subtraction images and presence of capsular appearance on portal venous or transitional phases conventional and subtraction images. We assessed effect of capsular appearance on subtraction images on HCC. Results: After excluding 1 patient who had treated by transarterial chemoembolization prior to surgery and 33 patients with unsatisfactory subtraction image qualities, 82 focal hepatic lesions (73 HCC, 5 non-HCC malignancies, and 4 benign) from 74 patients were analyzed. Regarding detection of capsules, sensitivity, accuracy, and area under the receiver operating characteristic curve (AUC) on subtraction images were significantly higher than those on conventional images (95.4%, 89.0%, and 0.80, respectively; p < 0.001), though specificities were same (64.7%). For diagnosis of HCC, sensitivity, accuracy, and AUC on subtraction images were significantly higher than on conventional images (82.2%, 79.3%, and 0.69, respectively; p = 0.011), though specificities were identical (55.6%). Conclusion: Portal venous or transitional phase gadoxetic acid-enhanced MRI subtraction images could improve detection of HCC capsule.

• Journal of Korea Multimedia Society
• /
• v.24 no.2
• /
• pp.178-185
• /
• 2021

Three-dimensional high-resolution magnetic resonance imaging (MRI) provides fine-level anatomical information for disease diagnosis. However, there is a limitation in obtaining high resolution due to the long scan time for wide spatial coverage. Therefore, in order to obtain a clear high-resolution(HR) image in a wide spatial coverage, a super-resolution technology that converts a low-resolution(LR) MRI image into a high-resolution is required. In this paper, we propose a super-resolution technique through filter learning based on information on the surrounding gradient information in 3D space from 3D MRI images. In the learning step, the gradient features of each voxel are computed through eigen-decomposition from 3D patch. Based on these features, we get the learned filters that minimize the difference of intensity between pairs of LR and HR images for similar features. In test step, the gradient feature of the patch is obtained for each voxel, and the filter is applied by selecting a filter corresponding to the feature closest to it. As a result of learning 100 T1 brain MRI images of HCP which is publicly opened, we showed that the performance improved by up to about 11% compared to the traditional interpolation method.