• The Korean Journal of Nuclear Medicine Technology
• /
• v.13 no.1
• /
• pp.57-62
• /
• 2009

Purpose: Brain SPECT study is more sensitive to motion than other studies. Especially, when applying 1-day subtraction method for Diamox SPECT, it needs shorter study time in order to prevent reexamination. We were required to have new study condition and analysing method on dual detector system because triple head camera in Seoul National University Hospital is to be disposed. So we have tried to increase image quality and make the dual and triple head to have equivalent study time by using a new analysing program. Materials and Methods: Using IEC phantom, we estimated contrast, SNR and FWHM. In Hoffman 3D brain phantom which is similar with real brain, we were on the supposition that 5% of injected doses were distributed in brain tissue. To compare with existing FBP method, we used fan-beam collimator. And we applied 15 sec, 25 sec/frame for each SEPCT studies using LEHR and LEUHR. We used OSEM2D and Onco-flash3D reconstruction method and compared reconstruction methods between applied Gaussian post-filtering 5mm and not applied as well. Attenuation correction was applied by manual method. And we did Brain SPECT to patient injected 15 mCi of $^{99m}Tc$-HMPAO according to results of Phantom study. Lastly, technologist, MD, PhD estimated the results. Results: The study shows that reconstruction method by Flash3D is better than exiting FBP and OSEM2D when studied using IEC phantom. Flowing by estimation, when using Flash3D, both of 15 sec and 25 sec are needed postfiltering 5 mm. And 8 times are proper for subset 8 iteration in Flash3D. OSEM2D needs post-filtering. And it is proper that subset 4, iteration 8 times for 15sec and subset 8, iteration 12 times for 25sec. The study regarding to injected doses for a patient and study time, combination of input parameter-15 sec/frame, LEHR collimator, analysing program-Flash3D, subset 8, iteration 8times and Gaussian post-filtering 5mm is the most appropriate. On the other hands, it was not appropriate to apply LEUHR collimator to 1-day subtraction method of Diamox study because of lower sensitivity. Conclusions: We could prove that there was also an advantage of short study time effectiveness in Dual camera same as Triple gamma camera and get great result of alternation from existing fan-beam collimator to parallel collimator. In addition, resolution and contrast of new method was better than FBP method. And it could improve sensitivity and accuracy of image because lesser subjectivity was input than Metz filter of FBP. We expect better image quality and shorter study time of Brain SPECT on Dual detector system.

• Journal of Periodontal and Implant Science
• /
• v.32 no.1
• /
• pp.199-211
• /
• 2002

Digital substraction technique and computer-assisted densitometirc analysis detect minor change in bone density and thus increase the diagnostic accuracy. This advantage as well as high sensitivity and objectivity which precludes human bias have drawn interest in radiologic research area. The objectives of this study are to verify if Radiographic density can be recognized in linear pattern when density profile of standard periapical radiograph with the aluminium stepwedge as the reference, was investigated under varies circumstances which can be encountered in clinical situations, and in addition to that to obtain mutual relationship between the existing standard radiographic system, and future digital image systems, by confirming the corelationship between the standard radiograph and Digora system which is a digital image system currently being used. In order to make quantitative analysis of the bone tissue, digital image system which uses high resolution automatic slide scanner as an input device, and Digora system were compared and analyzed using multifunctional program, Brain3dsp. The following conclusions were obtained. 1. Under common clinical situation that is 70kVp, 0.2 sec., and focal distance 10cm, Al-Equivalent image equation was found to be Y=11.21X+46.62 $r^2=0.9898$ in standard radiographic system, and Y=12.68X+74.59, $r^2=0.9528$ in Digora system, and linear relation was confirmed in both the systems. 2. In standard radiographic system, when all conditions were maintained the same except for the condition of developing solution, Al-Equivalent image equation was Y=10.07X+41.64, $r^2=0.9861$ which shows high corelationship. 3. When all conditions were maintained the same except for the Kilovoltage peak, linear relationship was still maintained under 60kVp, and Al-Equivalent image equation was Y=14.60X+68.86, $r^2=0.9886$ in the standard radiograhic system, and Y=13.90X+80.68, $r^2=0.9238$ in Digora system. 4. When all conditions were maintained the same except for the exposure time which was varied from 0.01 sec. to 0.8 sec., Al-Equivalent image equation was found to be linear in both the standard radiographic system and Digora system. The R-square was distributed from 0.9188 to 0.9900, and in general, standard radiographic system showed higher R-square than Digora system. 5. When all conditions were maintained the same except for the focal distance which was varied from 5cm to 30cm, Al-Equivalent image equation was found to be linear in both the standard radiographic system and Digora system. The R-square was distributed from 0.9463 to 0.9925, and the standard radiographic system had the tendency to show higher R-square in shorter focal distances.

• Journal of the Korea Computer Graphics Society
• /
• v.16 no.1
• /
• pp.1-7
• /
• 2010

With the widespread use of vision-based surveillance systems, the capability for person identification is now an essential component. However, the CCTV cameras used in surveillance systems tend to produce relatively low-resolution images, making it difficult to use face recognition techniques for person identification. Therefore, an algorithm is proposed for person identification in CCTV camera images based on the clothing. Whenever a person is authenticated at the main entrance of a building, the clothing feature of that person is extracted and added to the database. Using a given image, the clothing area is detected using background subtraction and skin color detection techniques. The clothing feature vector is then composed of textural and color features of the clothing region, where the textural feature is extracted based on a local edge histogram, while the color feature is extracted using octree-based quantization of a color map. When given a query image, the person can then be identified by finding the most similar clothing feature from the database, where the Euclidean distance is used as the similarity measure. Experimental results show an 80% success rate for person identification with the proposed algorithm, and only a 43% success rate when using face recognition.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.27 no.5
• /
• pp.426-431
• /
• 2007

When a high-energy ultrasound propagates through a solid body that contains a crack or a delamination, the two faces of the defect do not ordinarily vibrate in unison, and dissipative phenomena such as friction, rubbing and clapping between the faces will convert some of the vibrational energy to heat. By combining this heating effect with infrared imaging, one can detect a subsurface defect in material in real time. In this paper a realtime detection of the brazing defect of thin Inconel plates using the UIR (ultrasonic infrared imaging) technology is described. A low frequency (23 kHz) ultrasonic transducer was used to infuse the welded Inconel plates with a short pulse of sound for 280 ms. The ultrasonic source has a maximum power of 2 kW. The surface temperature of the area under inspection is imaged by an infrared camera that is coupled to a fast frame grabber in a computer. The hot spots, which are a small area around the bound between the two faces of the Inconel plates near the defective brazing point and heated up highly, are observed. And the weak thermal signal is observed at the defect position of brazed plate also. Using the image processing technology such as background subtraction average and image enhancement using histogram equalization, the position of defective brazing regions in the thin Inconel plates can be located certainly.

• Journal of Periodontal and Implant Science
• /
• v.37 no.1
• /
• pp.137-150
• /
• 2007

CADIA(Computer-assisted densitometric image analysis) method is used to analyze bone density changes around the implants. The usefullness and reproducibility of the method was assessed. We tried to find out if there is any possibility to quantitiate and qualitify peri-implant bone density change as time passes. And we concluded that this newly developed linear analysis is efficient for analyzing peri-implant bone density change non-Invasively. In this study, 2152 machined $Br{\aa}nemark$ fixtures installed from 1994 to 2002 in the department of Periodontics, Dental hospital of College of Dentistry, Yonsei University were included. Of these fixtures 22 radiographically analyzable failed fixtures were used as experimental group, and 22 successful implants placed in the same patient were used as control group. 1. 57 out of 1635 machined $Br{\aa}nemark$ standard and Mk II implants system failed, the survival rate was 96.5%. And 11 out of 517 machined $Br{\aa}nemark$ Mk III and Mk IV implants system failed, the survival rate was 97.9%. Total survival rate was 96.8%. 2. 22 failed implants were used for the analysis, 10 of which failed before prosthetic treatment due to infection and overheating. 12 failed due to overload after prosthetic treatment, 63.6% of which failed during the early phase of functional loading, i, e. before 1 year of loading. 3. Bone density change values around coronal region of the failed implants were $-6.54{\pm}6.35$, middle region were $-3.53{\pm}5.78$, apical region were $-0.75{\pm}10.33$, resulting in average of $-3.71{\pm}8.03$. 4. Bone density change values around coronal region of the successful implants were $4.25{\pm}4.66$, middle region were $6.33{\pm}5.02$, apical region were $9.89{\pm}4.67$, resulting in average of $6.27{\pm}5.29$. 5. There was a statistically significant difference between two groups (p<0.01). In conclusion, the linear analysis method using computer-assisted densitometric image analysis could be a useful method for the analysis of implants, and could be used for future implant researchs.

• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.139-156
• /
• 2021

The biggest reason for using a deep learning model in image classification is that it is possible to consider the relationship between each region by extracting each region's features from the overall information of the image. However, the CNN model may not be suitable for emotional image data without the image's regional features. To solve the difficulty of classifying emotion images, many researchers each year propose a CNN-based architecture suitable for emotion images. Studies on the relationship between color and human emotion were also conducted, and results were derived that different emotions are induced according to color. In studies using deep learning, there have been studies that apply color information to image subtraction classification. The case where the image's color information is additionally used than the case where the classification model is trained with only the image improves the accuracy of classifying image emotions. This study proposes two ways to increase the accuracy by incorporating the result value after the model classifies an image's emotion. Both methods improve accuracy by modifying the result value based on statistics using the color of the picture. When performing the test by finding the two-color combinations most distributed for all training data, the two-color combinations most distributed for each test data image were found. The result values were corrected according to the color combination distribution. This method weights the result value obtained after the model classifies an image's emotion by creating an expression based on the log function and the exponential function. Emotion6, classified into six emotions, and Artphoto classified into eight categories were used for the image data. Densenet169, Mnasnet, Resnet101, Resnet152, and Vgg19 architectures were used for the CNN model, and the performance evaluation was compared before and after applying the two-stage learning to the CNN model. Inspired by color psychology, which deals with the relationship between colors and emotions, when creating a model that classifies an image's sentiment, we studied how to improve accuracy by modifying the result values based on color. Sixteen colors were used: red, orange, yellow, green, blue, indigo, purple, turquoise, pink, magenta, brown, gray, silver, gold, white, and black. It has meaning. Using Scikit-learn's Clustering, the seven colors that are primarily distributed in the image are checked. Then, the RGB coordinate values of the colors from the image are compared with the RGB coordinate values of the 16 colors presented in the above data. That is, it was converted to the closest color. Suppose three or more color combinations are selected. In that case, too many color combinations occur, resulting in a problem in which the distribution is scattered, so a situation fewer influences the result value. Therefore, to solve this problem, two-color combinations were found and weighted to the model. Before training, the most distributed color combinations were found for all training data images. The distribution of color combinations for each class was stored in a Python dictionary format to be used during testing. During the test, the two-color combinations that are most distributed for each test data image are found. After that, we checked how the color combinations were distributed in the training data and corrected the result. We devised several equations to weight the result value from the model based on the extracted color as described above. The data set was randomly divided by 80:20, and the model was verified using 20% of the data as a test set. After splitting the remaining 80% of the data into five divisions to perform 5-fold cross-validation, the model was trained five times using different verification datasets. Finally, the performance was checked using the test dataset that was previously separated. Adam was used as the activation function, and the learning rate was set to 0.01. The training was performed as much as 20 epochs, and if the validation loss value did not decrease during five epochs of learning, the experiment was stopped. Early tapping was set to load the model with the best validation loss value. The classification accuracy was better when the extracted information using color properties was used together than the case using only the CNN architecture.

• Journal of Korea Society of Industrial Information Systems
• /
• v.17 no.4
• /
• pp.61-70
• /
• 2012

We propose a method that finds candidate targets based on morphology and detects a small target from them using modified gaussian distance function. The existing small target detection methods use predictive filters or morphology. The methods using predictive filters take long to approach least errors. The methods using morphology are weak at clutters and need to consider size of a small target when selecting size of structure elements. We propose a robust method for small target detection to complete the existing methods. First, the proposed method deletes clutters using a median filter. Next, it does closing and opening operation using various size of structure elements, and figures target candidate pixels with subtraction operation between the results of closing and opening operation. It detects an exact small target using a gaussian distance function from the candidates target areas. The proposed method is less sensitive to clutters, and shows a detection rate of 98%.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.671-674
• /
• 2005

Atmospheric correction over the ocean part is more important than that over the land because the signal from the ocean is very small about one tenth of that reflected from land. In this study, the Spectral Shape Matching Method (SSMM) developed by Ahn and Shanmugam (2004) is evaluated using Landsat imagery acquired over the highly turbid Saemangeum Coastal Area. The result of SSMM is compared with COST model developed by Chavez (1991 and 1997). In principle, SSMM is simple and easy to implement on any satellite imagery, relying on both field and image properties. To assess the potential use of these methods, several field campaigns were conducted in the Saemangeum coastal area corresponding with Landsat-7 satellite's overpass on 29 May 2005. In-situ data collected from the coastal waters of Saemangeum using optical instruments (ASD field spectroradiometer) consists of ChI, Ap, SS, aooM, F(d). In order to perform SSMM, we use the in-situ water-leaving radiance spectra from clear oceanic waters to estimate the the path radiance from total signal recorded at the top of the atmosphere (TOA), due to the reason that the shape of clear water-leaving radiance spectra is nearly stable than turbid water-leaving radiance spectra. The retrieved water-leaving radiance after subtraction of path signal from TOA signal in this way is compared with that estimated by COST model. The result shows that SSMM enabled retrieval of water-leaving radiance spectra that are consistent with in-situ data obtained from Saemangeum coastal waters. The COST model yielded significantly high errors in these areas.

• Journal of Korean Neurosurgical Society
• /
• v.53 no.4
• /
• pp.235-240
• /
• 2013

Fusiform aneurysms on the basilar artery (BA) trunk are rare. The microsurgical management of these aneurysms is difficult because of their deep location, dense collection of vital cranial nerves, and perforating arteries to the brain stem. Endovascular treatment is relatively easier and safer compared with microsurgical treatment. Selective occlusion of the aneurysmal sac with preservation of the parent artery is the endovascular treatment of choice. But, some cases, particularly giant or fusiform aneurysms, are unsuitable for selective sac occlusion. Therefore, endovascular coiling of the aneurysm with parent vessel occlusion is an alternative treatment option. In this situation, it is important to determine whether a patient can tolerate parent vessel occlusion without developing neurological deficits. We report a rare case of fusiform aneurysms in the BA trunk. An 18-year-old female suffered a headache for 2 weeks. Computed tomography and magnetic resonance image revealed a fusiform aneurysm of the lower basilar artery trunk. Digital subtraction angiography revealed a $7.1{\times}11.0$ mm-sized fusiform aneurysm located between vertebrovasilar junction and the anterior inferior cerebellar arteries. We had good clinical result using endovascular coiling of unruptured fusiform aneurysm on the lower BA trunk with parent vessel occlusion after confirming the tolerance of the patient by balloon test occlusion with induced hypotension and accompanied by neurophysiologic monitoring, transcranial Doppler and single photon emission computed tomography. In this study, we discuss the importance of preoperative meticulous studies for avoidance of delayed neurological deficit in the patient with fusiform aneurysm on lower basilar trunk.

• The Journal of the Korea Contents Association
• /
• v.18 no.2
• /
• pp.1-10
• /
• 2018

Aggression among pigs adversely affects economic returns and animal welfare in intensive pigsties. Recently, some studies have applied information technology to a livestock management system to minimize the damage resulting from such anomalies. Nonetheless, detecting each pig in a crowed pigsty is still challenging problem. In this paper, we propose a new Kinect camera and deep learning-based monitoring system for the detection of the individual pigs. The proposed system is characterized as follows. 1) The background subtraction method and depth-threshold are used to detect only standing-pigs in the Kinect-depth image. 2) The standing-pigs are detected by using YOLO (You Only Look Once) which is the fastest and most accurate model in deep learning algorithms. Our experimental results show that this method is effective for detecting individual pigs in real time in terms of both cost-effectiveness (using a low-cost Kinect depth sensor) and accuracy (average 99.40% detection accuracies).