• 대한의용생체공학회:의공학회지
• /
• 제42권4호
• /
• pp.150-158
• /
• 2021

Intracranial hemorrhage (ICH) refers to acute bleeding inside the intracranial vault. Not only does this devastating disease record a very high mortality rate, but it can also cause serious chronic impairment of sensory, motor, and cognitive functions. Therefore, a prompt and professional diagnosis of the disease is highly critical. Noninvasive brain imaging data are essential for clinicians to efficiently diagnose the locus of brain lesion, volume of bleeding, and subsequent cortical damage, and to take clinical interventions. In particular, computed tomography (CT) images are used most often for the diagnosis of ICH. In order to diagnose ICH through CT images, not only medical specialists with a sufficient number of diagnosis experiences are required, but even when this condition is met, there are many cases where bleeding cannot be successfully detected due to factors such as low signal ratio and artifacts of the image itself. In addition, discrepancies between interpretations or even misinterpretations might exist causing critical clinical consequences. To resolve these clinical problems, we developed a diagnostic model predicting intracranial bleeding and its subtypes (intraparenchymal, intraventricular, subarachnoid, subdural, and epidural) by applying deep learning algorithms to CT images. We also constructed a visualization tool highlighting important regions in a CT image for predicting ICH. Specifically, 1) 27,758 CT brain images from RSNA were pre-processed to minimize the computational load. 2) Three different CNN-based models (ResNet, EfficientNet-B2, and EfficientNet-B7) were trained based on a training image data set. 3) Diagnosis performance of each of the three models was evaluated based on an independent test image data set: As a result of the model comparison, EfficientNet-B7's performance (classification accuracy = 91%) was a way greater than the other models. 4) Finally, based on the result of EfficientNet-B7, we visualized the lesions of internal bleeding using the Grad-CAM. Our research suggests that artificial intelligence-based diagnostic systems can help diagnose and treat brain diseases resolving various problems in clinical situations.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2006년도 춘계학술대회 논문집
• /
• pp.171-172
• /
• 2006

Most existing studies for stress analysis of teeth have employed small number of teeth, and used big element size using F-E models. Therefore, the results are not accuracy enough for showing the internal stress variation. 15 males' and 13 females' mandibular first premolar are employed for internal structure's study of teeth and small element size for a FE model are used. According to these processes, stress distribution of internal parts of teeth are well shown, and the stresses are varied a lot between enamel layer and dentine layer, but there is little variation on pulp chamber.

• 대한방사선기술학회지:방사선기술과학
• /
• 제44권6호
• /
• pp.645-653
• /
• 2021

With the recent development of precision medicine(Theranostics), interest and utilization of the quantitative function of SPECT/CT are increasing. This study aims to investigate the effect on the radioactivity concentration estimate by the increase or decrease in the total time of SPECT/CT imaging conditions. A standard image was obtained by the conditions of a total acquisition time of 600 sec(10 sec/f × 120 frames) by diluting ^99mTc 91.76 MBq in a cylindrical phantom filled with sterile water, and a comparative image was obtained by increasing the total acquisition time by -90%, -75%, -50%, -25%, +50%, +100%. The CNR, radioactive concentration estimate(cps/ml), and the variation rate(%) of the recovery coefficient(RC) were analyzed by measuring the overall coefficient of interest in each image. The results[CNR, Radiation Concentration, RC] by the change in the number of projections for each increase or decrease rate(-90%, -75%, -50%, -25%, +50%, +100%) of total acquisition time are as follows. [-89.5%, +3.90%, 1.04] at -90%, [-77.9%, +2.71%, 1.03] at -75%, [-55.6%, +1.85%, 1.02] at -50%, [-33.6%, +1.37%, 1.01] at -25%, [-33.7%, +0.71%, 1.01] at +50%, [+93.2%, +0.32%, 1.00] at +100%. and also The results[CNR, Radiation Concentration, RC] by the acquisition time change for each increase or decrease rate(-90%, -75%, -50%, -25%, +50%, +100%) of total acquisition time are as follows. [-89.3%, -3.55%, 0.96] at -90%, [-73.4%, -0.17%, 1.00] at -75%, [-49.6%, -0.34%, 1.00] at -50%, [-24.9%, 0.03%, 1.00] at -25%, [+49.3%, -0.04%, 1.00] at +50%, [+99.0%, +0.11%, 1.00] at +100%. Image quality(CNR) showed a pattern of change in proportion to the increase or decrease in the total acquisition time of SPECT/CT, but the result at quantitative evaluation showed a change of less than 5% in all experimental conditions, maintaining quantitative accuracy(RC less than 0.05) without much influence.

• 핵의학기술
• /
• 제14권1호
• /
• pp.13-17
• /
• 2010

PVE는 PET/CT 3D 영상 획득에서 발생되는 것으로 평가값이 저평가되어 영상의 정확도를 떨어뜨리는 현상이다. 특히 이는 병소의 크기가 작고 분해능이 저하될수록 더 큰 오차를 초래하여 검사 결과에 영향을 줄 수 있다. 본 연구는 PVE에 영향을 줄 수 있는 매개변수의 변화를 이용하여 최적의 영상 재구성법을 알아보고자 한다. GE Discovery STE16 장비에서 NEMA 2001 IEC phantom을 이용하여 각기 다른 크기의 구체(직경 37, 28, 22, 17, 13, 10 mm)에 $^{18}F$-FDG를 열소와 배후방사능비 4:1로 주입하여 10분간 영상을 획득하였다. 재구성은 반복재구성법(iterative reconstruction)을 사용하였으며, 반복 횟수(iteration) 2~50회, 부분집합 수(subset number) 1~56개로 변화를 주었다. 분석은 영상의 구체부분에 관심영역(ROI)을 설정하고 최대 표준섭취계수($SUV_{max}$)를 이용하여 백분율 차이(% difference)와 신호대잡음비(SNR)를 산출하였다. 반복 횟수 2, 6, 13, 30, 50회 변화를 준 10 mm 구체의 $SUV_{max}$는 2.32, 3.60, 3.88, 3.88, 3.90이고, SNR은 0.36, 0.49, 0.47, 0.43, 0.41이었으며, 백분율 차이는 58.9, 38.5, 34.8, 35.7, 35.4%로 측정되었다. 또한 6회로 고정한 반복 횟수에 2, 5, 8, 20, 56으로 부분집합 수를 변화시킨 평균 $SUV_{max}$는 10mm의 구체에서 1.46, 3.10, 3.10, 3.48, 3.78로 측정되었으며, SNR은 0.19, 0.30, 0.40, 0.48, 0.45로 나타났다. 또한 각 구체의 SNR의 합은 2.73, 3.38, 3.64, 3.63, 3.38로 측정되었다. 반복 횟수 6회부터 20회까지는 평균 백분율 차이($73{\pm}1%$)와 평균 SNR ($3.47{\pm}0.09$)은 비슷한 값을 나타내었으며, 20회 이상에서는 noise의 영향으로 SUV가 저평가되는 현상이 증가하였다. 또한 동일한 반복 횟수의 경우에 부분집합 값의 변화에서 SNR은 8회부터 20회가 높은 구간($3.63{\pm}0.002$)으로 나타났다. 따라서 작은 병소의 PVE를 줄이기 위해서는 재구성 시간을 고려하여 반복 횟수 6회, 부분집합 수 8~20회에서 PVE를 가장 저감할 수 있다.

• 대한핵의학회지
• /
• 제29권4호
• /
• pp.465-472
• /
• 1995

목적 : 원발성 부갑상선기능 항진증의 대부분의 원인인 부갑상선 선종의 유일한 완치는 수술이며 수술의 성패에 가장 중요한 것은 수술전의 정확한 ??소의 확인이다. 다양한 영상검사들이 동원되고있으나 각가 결점이 있고 가장 특이도가 높은 $^{201}T1-^{99m}Tc$감영 촬영방법도 검사방법등에 문제점이 있다. 이에 원발성 부갑상선기능 항진증 환자에게 최근에 사용되고있는 $^{99m}Tc$-Sestamibi 부갑상선 스캔의 유용성을 알아보고자 하였다. 대상 및 방법 : 19명의 부갑상선기능 항진증환자에서 모든 환자에게 초음파 검사, CT, $^{201}T1-^{99m}Tc$감영 촬영을 시행하고 그중 5명은 $^{99m}Tc$-Sestamibi 부갑상선 스캔을 시행하여 수술후 확진결과와 함께 비교하였다. 결과 : 초음파 검사와 CT 및 $^{201}T1-^{99m}Tc$감영 스캔은 19예중 각가 12예(67.1%), 14예(78.9%) 그리고 17예(89.5%)에서 병소확인이 가능하였고 $^{99m}Tc$-Sestamibi 부갑상선 스캔은 시행한 5예(100%) 모두에서 병소 확인이 가능하였다. 결론 : 이상의 결과로 $^{99m}Tc$-Sestamibi 부갑상선 스캔은 원발성 부갑상선기능 항진증의 수술전 병소확인에서 유용한 검사법으로 이용될 수 있으리라 사료된다.

• Journal of Chest Surgery
• /
• 제31권7호
• /
• pp.730-734
• /
• 1998

배경: 기흉 환자를 수술하기 전에 폐기포의 위치, 숫자, 크기와 모양을 파악하기 위하여 단순 흉부 X선 사진이나 흉부 컴퓨터 단층 촬영등을 이용한다. 단순 흉부 X선 사진은 간편하고 저렴하나 정확도가 낮고, 흉부 컴퓨터 단층 촬영은 정확하나 비용이 많이든다. 그러나 흉강조영술을 이용하면, 폐기포의 상태를 비교적 저렴하고 정확히 파악할수 있으리라 예상되어, 이를 확인하기 위하여 본 임상 실험을 계획하였다. 대상 및 방법: 대상은 개흉이나 흉강내시경을 이용하여 폐기포 절제를 시행한, 원발성 자연 기흉 환자 22명이었다. 수술전에 단순 흉부 X선 사진과 흉강조영술을 시행하여, 이를 수술시의 폐기포의 소견과 비교하였다. 결과: 흉강조영술의 Sensitivity는 75%로 단순 흉부 X선 사진의 30%에 비하여 높았고, 정확도(Accuracy)도 72.7%로 단순 흉부 X선 사진의 36.4%에 비하여 높았다. 그러나 Specificity는 흉강조영술이 50%로 단순 흉부 X선 사진의 100%에 비하여 낮았다. 검사로인한 합병증은 없었다. 결론: 결론적으로 기흉의 수술전 검사로서 흉강조영술은 단순 흉부 X선 사진에 비하여 높은 진단적 가치를 지녔으며, 안전하게 이용될 수 있으리라 생각된다.

• Journal of Korean Neurosurgical Society
• /
• 제30권sup2호
• /
• pp.294-299
• /
• 2001

Objectives : CT-guided stereotactic evacuation for spontaneous intracerebral hemorrhage can minimize the brain damage and can be performed safely and simply under local anesthesia. But that procedure is time consuming and has a risk of rebleeding because of the stress during head pin fixation. So authors describe easy and precise guidelines for FHA of putaminal hemorrhage without stereotactic instrument. Methods and Materials : We analyzed the data of 298 patients who underwent CT-guided stereotactic aspiration of putaminal hematoma in our hospital between January 1990 and December 2000. We divided the patients into three groups according to the location of hematoma : anterior portion, middle portion and posterior portion of putamen. Total number of catheters inserted into the hematoma were 345 and there were with regard to the direction and depth of catheters. Results : Proposed guidelines of catheter insertion to putaminal hemorrhage in our institution. 1) hematoma at the anterior portion of putamen ; Direction of catheter was the midpupillary line of the eye and the point intersecting a line drawn from the burr hole to a point between external auditory meatus(EOM) and 1cm posterior to EOM. Depth of catheter was 6-6.5cm. 2) hematoma at the middle portion of putamen ; Direction of catheter was the midpupillary line of the the eye and the point intersecting a line drawn from the burr hole to a point between 1cm and 2cm posterior to EOM. Depth of catheter was 6.5-7cm. 3) hematoma at the posterior portion of putamen ; Direction of catheter was 15 degree laterally from the midpupillary line of the eye and the point intersecting a line drawn from the burr hole to a point between 2cm and 3cm posterior to EOM. Depth of catheter was 7-7.5cm. We have performed FHA of putaminal hemorrhage in 48 cases according to this guideline. All catheter were inserted exactly at the center of hematoma and average operation time was about 30 minutes. Conclusion : Our proposed guidelines for putaminal hemorrhage are considered to be safe and simple method with similar accuracy and rapid decompression compared with traditional stereotactic method. Main advantages of this technique were unnecessity of stereotactic frame application and less time requirement for hematoma removal.

• Korean Journal of Radiology
• /
• 제21권6호
• /
• pp.660-669
• /
• 2020

Objective: To evaluate the accuracy of a deep learning-based automated segmentation of the left ventricle (LV) myocardium using cardiac CT. Materials and Methods: To develop a fully automated algorithm, 100 subjects with coronary artery disease were randomly selected as a development set (50 training / 20 validation / 30 internal test). An experienced cardiac radiologist generated the manual segmentation of the development set. The trained model was evaluated using 1000 validation set generated by an experienced technician. Visual assessment was performed to compare the manual and automatic segmentations. In a quantitative analysis, sensitivity and specificity were calculated according to the number of pixels where two three-dimensional masks of the manual and deep learning segmentations overlapped. Similarity indices, such as the Dice similarity coefficient (DSC), were used to evaluate the margin of each segmented masks. Results: The sensitivity and specificity of automated segmentation for each segment (1-16 segments) were high (85.5-100.0%). The DSC was 88.3 ± 6.2%. Among randomly selected 100 cases, all manual segmentation and deep learning masks for visual analysis were classified as very accurate to mostly accurate and there were no inaccurate cases (manual vs. deep learning: very accurate, 31 vs. 53; accurate, 64 vs. 39; mostly accurate, 15 vs. 8). The number of very accurate cases for deep learning masks was greater than that for manually segmented masks. Conclusion: We present deep learning-based automatic segmentation of the LV myocardium and the results are comparable to manual segmentation data with high sensitivity, specificity, and high similarity scores.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제35권2호
• /
• pp.118-127
• /
• 2008

본 논문에서는 하악골 데이타에서 견고한 통계 형상 모델을 생성하기 위한 방법을 제안한다. 본 제안 방법은 다음과 같은 네 단계로 구성된다. 첫째, 3차원 입력 형상에 대해 패치 분할을 수행한다. 둘째, 부동 형상에 대한 대응 형상 생성을 위하여 훈련 집합의 모든 형상들을 패치의 형태와 비슷한 도형인 원에 매개화 과정을 수행한다. 셋째, 기준 형상과 각 부동 형상 간 일대일 매핑을 통해 대응 형상을 생성한다. 이때, 패치 경계 부분에서 대응 정점 생성이 불가능한 문제를 해결한다. 마지막으로 대응 형상들을 기준 형상으로 정렬하고, 주성분 분석 기법을 사용하여 통계 형상 모델을 생성한다. 제안 방법을 적용하여 생성한 3차원 하악골 통계 형상 모델의 정확성을 평가하기 위해 육안 평가와 부동 형상과 대응 형상 간 평균 거리 차이를 이용한 유사성 측정을 수행한다. 또한 형상 변화를 표현하는 모드를 이용하여 통계 형상 모델의 밀집도를 측정한다. 실험 결과 다양한 특성을 갖는 하악골 데이타로 생성된 3차원 통계 형상 모델은 부동 형상과 대응 형상 간 높은 유사성을 가지며 적은 수의 모드로 통계 형상 모델 표현됨을 보여 준다.

• 치과방사선
• /
• 제28권1호
• /
• pp.225-234
• /
• 1998

We used five adult dog mandibles embedded in resin block and six different cross-sectional planes for each mandible were choosen. According to the angle of mandibular occulsal plane to vertical plane(mandibular angle) and gantry angle of CT machine, we classified 4 experimental groups and 1 control group. The control group images were taken at the mandibular angel 0° and gantry angle 0°. The experimental images were taken at the mandibular angle 15° and gantry angle 0°(group 1); 30° and 0°(group 2); 15° and 15°(group 3) ;30° and 30°(group 4), respectively. Using the reformatted cross-sectional images, the distance from the mandibular canal to the alveolar crest and the distance from the mandibular canal to the buccal cortex and to the lingual cortex was measured and compared. The obtained results were as follows: 1. The distance from the mandibular canal to the alveolar crest of group 1 and 2 was larger than control group, but the distance of group 3 and 4 was smaller. The distance from the mandibular canal to the buccal cortex and to the lingual cortex of all experimental groups was smaller than control group. 2. The distance from the mandibular canal to the alveolar crest showed the largest difference from control group in all experimental groups, especially in group 2 and 4(p<0.05). 3. In the distance from the mandibular canal to the alveolar crest, the number of deviation value under 1 mm was 20 in group 3 and was 11 in group 2 and 4, respectively. 4. The deviation value of the distance from the mandibular canal to the buccal cortex and to the lingual cortex was under 1 mm in most cases.