• 대한의용생체공학회:의공학회지
• /
• 제44권4호
• /
• pp.255-263
• /
• 2023

Clinical ultrasound (US) is a widely used imaging modality with various clinical applications. However, capturing a large field of view often requires specialized transducers which have limitations for specific clinical scenarios. Panoramic imaging offers an alternative approach by sequentially aligning image sections acquired from freehand sweeps using a standard transducer. To reconstruct a 3D volume from these 2D sections, an external device can be employed to track the transducer's motion accurately. However, the presence of optical or electrical interferences in a clinical setting often leads to incorrect measurements from such sensors. In this paper, we propose a deep learning (DL) framework that enables the prediction of scan trajectories using only US data, eliminating the need for an external tracking device. Our approach incorporates diverse data types, including correlation volume, optical flow, B-mode images, and rawer data (IQ data). We develop a DL network capable of effectively handling these data types and introduce an attention technique to emphasize crucial local areas for precise trajectory prediction. Through extensive experimentation, we demonstrate the superiority of our proposed method over other DL-based approaches in terms of long trajectory prediction performance. Our findings highlight the potential of employing DL techniques for trajectory estimation in clinical ultrasound, offering a promising alternative for panoramic imaging.

• Journal of Korean Neurosurgical Society
• /
• 제60권4호
• /
• pp.441-447
• /
• 2017

Objective : Computed tomography (CT)-based method of three dimensional (3D) analysis ($MIMICS^{(R)}$, Materialise, Leuven, Belgium) is reported as very useful software for evaluation of OPLL, but its reliability and reproducibility are obscure. This study was conducted to evaluate the accuracy of $MIMICS^{(R)}$ system, and inter- and intra-observer reliability in the measurement of OPLL. Methods : Three neurosurgeons independently analyzed the randomly selected 10 OPLL cases with medical image processing software ($MIMICS^{(R)}$) which create 3D model with Digital Imaging and Communication in Medicine (DICOM) data from CT images after brief explanation was given to examiners before the image construction steps. To assess the reliability of inter- and intra-examiner intraclass correlation coefficient (ICC), 3 examiners measured 4 parameters (volume, length, width, and length) in 10 cases 2 times with 1-week interval. Results : The inter-examiner ICCs among 3 examiners were 0.996 (95% confidence interval [CI], 0.987-0.999) for volume measurement, 0.973 (95% CI, 0.907-0.978) for thickness, 0.969 (95% CI, 0.895-0.993) for width, and 0.995 (95% CI, 0.983-0.999) for length. The intra-examiner ICCs were 0.994 (range, 0.991-0.996) for volume, 0.996 (range, 0.944-0.998) for length, 0.930 (range, 0.873-0.947) for width, and 0.987 (range, 0.985-0.995) for length. Conclusion : The medical image processing software ($MIMICS^{(R)}$) provided detailed quantification OPLL volume with minimal error of inter- and intra-observer reliability in the measurement of OPLL.

• 인터넷정보학회논문지
• /
• 제24권4호
• /
• pp.85-92
• /
• 2023

X-ray, Computed Tomography (CT), Magnetic Resonance Imaging (MRI)과 같은 의료데이터에서 딥러닝을 활용해 질병 유무 판별 태스크와 같은 문제를 해결하려는 시도가 활발하다. 대부분의 데이터 기반 딥러닝 문제들은 높은 정확도 달성과 정답과 비교하는 성능평가의 활용을 위해 지도학습기법을 사용해야 한다. 지도학습에는 다량의 이미지와 레이블 세트가 필요하지만, 학습에 충분한 양의 의료 이미지 데이터를 얻기는 어렵다. 다양한 데이터 증강 기법을 통해 적은 양의 의료이미지와 레이블 세트로 지도학습 기반 모델의 과소적합 문제를 극복할 수 있다. 본 연구는 딥러닝 기반 갈비뼈 골절 세그멘테이션 모델의 성능 향상과 효과적인 좌우 반전, 회전, 스케일링 등의 데이터 증강 기법을 탐색한다. 좌우 반전과 30° 회전, 60° 회전으로 증강한 데이터셋은 모델 성능 향상에 기여하지만, 90° 회전 및 ⨯0.5 스케일링은 모델 성능을 저하한다. 이는 데이터셋 및 태스크에 따라 적절한 데이터 증강 기법의 사용이 필요함을 나타낸다.

• 한국의학물리학회지:의학물리
• /
• 제28권1호
• /
• pp.27-32
• /
• 2017

We investigated the effect of a commercial iterative reconstruction technique (iDose, Philips) on the image quality and the dose calculation for the treatment plan. Using the electron density phantom, the 3D CT images with five different protocols (50, 100, 200, 350 and 400 mAs) were obtained. Additionally, the acquired data was reconstructed using the iDose with level 5. A lung phantom was used to acquire the 4D CT with the default protocol as a reference and the low dose (one third of the default protocol) 4D CT using the iDose for the spine and lung plans. When applying the iDose at the same mAs, the mean HU value was changed up to 85 HU. Although the 1 SD was increased with reducing the CT dose, it was decreased up to 4 HU due to the use of iDose. When using the low dose 4D CT with iDose, the dose change relative to the reference was less than 0.5% for the target and OARs in the spine plan. It was also less than 1.1% in the lung plan. Therefore, our results suggests that this dose reduction technique is applicable to the 4D CT image acquisition for the radiation treatment planning.

• The Journal of Advanced Prosthodontics
• /
• 제9권6호
• /
• pp.409-415
• /
• 2017

PURPOSE. Accurate information is essential in dentistry. The image information of missing teeth is used in optically based medical equipment in prosthodontic treatment. To evaluate oral scanners, the standardized model was examined from cases of image recognition errors of linear discriminant analysis (LDA), and a model that combines the variables with reference to ISO 12836:2015 was designed. MATERIALS AND METHODS. The basic model was fabricated by applying 4 factors to the tooth profile (chamfer, groove, curve, and square) and the bottom surface. Photo-type and video-type scanners were used to analyze 3D images after image capture. The scans were performed several times according to the prescribed sequence to distinguish the model from the one that did not form, and the results confirmed it to be the best. RESULTS. In the case of the initial basic model, a 3D shape could not be obtained by scanning even if several shots were taken. Subsequently, the recognition rate of the image was improved with every variable factor, and the difference depends on the tooth profile and the pattern of the floor surface. CONCLUSION. Based on the recognition error of the LDA, the recognition rate decreases when the model has a similar pattern. Therefore, to obtain the accurate 3D data, the difference of each class needs to be provided when developing a standardized model.

• 대한의용생체공학회:학술대회논문집
• /
• 대한의용생체공학회 1998년도 추계학술대회
• /
• pp.57-59
• /
• 1998

For medical students and doctors, knowledge of the three-dimensional (3D) structure of brain is very important in diagnosis and treatment of brain diseases. Two-dimensional (2D) tools (ex: anatomy book) or traditional 3D tools (ex: plastic model) are not sufficient to understand the complex structures of the brain. However, it is not always guaranteed to dissect the brain of cadaver when it is necessary. To overcome this problem, the virtual dissection programs of the brain have been developed. However, most programs include only 2D images that do not permit free dissection and free rotation. Many programs are made of radiographs that are not as realistic as sectioned cadaver because radiographs do not reveal true color and have limited resolution. It is also necessary to make the virtual dissection programs of each race and ethnic group. We attempted to make a virtual dissection program using a 3D image of the brain from a Korean cadaver. The purpose of this study is to present an educational tool for those interested in the anatomy of the brain. The procedures to make this program were as follows. A brain extracted from a 58-years old male Korean cadaver was embedded with gelatin solution, and serially sectioned into 1.4 mm-thickness using a meat slicer. 130 sectioned specimens were inputted to the computer using a scanner ($420\times456$ resolution, true color), and the 2D images were aligned on the alignment program composed using IDL language. Outlines of the brain components (cerebrum, cerebellum, brain stem, lentiform nucleus, caudate nucleus, thalamus, optic nerve, fornix, cerebral artery, and ventricle) were manually drawn from the 2D images on the CorelDRAW program. Multimedia data, including text and voice comments, were inputted to help the user to learn about the brain components. 3D images of the brain were reconstructed through the volume-based rendering of the 2D images. Using the 3D image of the brain as the main feature, virtual dissection program was composed using IDL language. Various dissection functions, such as dissecting 3D image of the brain at free angle to show its plane, presenting multimedia data of brain components, and rotating 3D image of the whole brain or selected brain components at free angle were established. This virtual dissection program is expected to become more advanced, and to be used widely through Internet or CD-title as an educational tool for medical students and doctors.

• 대한방사성의약품학회지
• /
• 제3권2호
• /
• pp.65-71
• /
• 2017

To assess the effects of filter and reconstruction of Cu-64 PET data on Siemens scanner, the various reconstruction algorithm with various filters were assessed in terms of spatial resolution, non-uniformity (NU), recovery coefficient (RC), and spillover ratio (SOR). Image reconstruction was performed using filtered backprojection (FBP), 2D ordered subset expectation maximization (OSEM), 3D reprojection algorithm (3DRP), and maximum a posteriori algorithms (MAP). For the FBP reconstruction, ramp, butterworth, hamming, hanning, or parzen filters were used. Attenuation or scatter correction were performed to assess the effect of attenuation and scatter correction. Regarding spatial resolution, highest achievable volumetric resolution was $3.08mm^3$ at the center of FOV when MAP (${\beta}=0.1$) reconstruction method was used. SOR was below 4% for FBP when ramp, Hamming, Hanning, or Shepp-logan filter were used. The lowest NU (highest uniform) after attenuation & scatter correction was 5.39% when FBP (parzen filter) was used. Regarding RC, 0.9 < RC < 1.1 was obtained when OSEM (iteration: 10) was used when attenuation and scatter correction were applied. In this study, image quality of Cu-64 on Siemens Inveon PET was investigated. This data will helpful for the quantification of Cu-64 PET data.

• 핵의학기술
• /
• 제13권1호
• /
• pp.9-14
• /
• 2009

최근 영상 처리 기법의 발전으로 영상의 질은 저하시키지 않고 검사 소요시간을 단축시키는 방법들이 개발되고 있다. 특히 단층 촬영의 경우 영상 재구성 방법을 개선하여 영상의 질이 우수한 영상을 획득할 수 있게 되었다. Philips사의 PRECEDENCE 16 감마카메라를 이용해 보편적으로 시행하고 있는 분석법에 의한 FBP 방법과 반복법에 의한 Astonish, 3D OSEM 방법을 이용해 각각 영상을 재구성하여 정성적인 분석과 정량적인 분석을 통해 영상 획득시간을 다르게 한 영상간의 비교와, 동일한 시간으로 획득한 영상을 비교하여 영상의 질이 우수한 재구성 방법에 대해 연구 하였다. 정성적인 분석을 위해 blind test를 한 결과, 영상 획득시간에 따른 영상의 질은 거의 차이가 없는 것을 확인할 수 있었다. 또한 정량적인 분석을 통해서도 영상 획득시간에 따른 영상의 질은 통계적으로 유의한 차이가 없었다. 하지만 영상 획득시간이 동일한 영상을 재구성 방법에 따라 분석한 결과는 통계적으로 유의한 차이가 있음을 확인할 수 있었다. 영상의 질은 반복법을 이용하는 Astonish에 의해 재구성된 영상이 해상력이 좋고 임상적으로 진단적 정보를 제공하는데 우수한 영상으로 판단된다. 영상을 재구성하기 위한 소요시간이 길고 저장 공간의 부족 등으로 현재까지 많이 사용되지 않던 반복법에 의한 재구성 방법이 영상의 질은 향상시키고 검사시간은 단축 할 수 있는 방법이 될 수 있음을 확인하였다.

• 한국의학물리학회지:의학물리
• /
• 제21권2호
• /
• pp.174-182
• /
• 2010

본 연구는 호흡 정보를 갖고 있는 PET 영상의 표준섭취계수(SUV: standard uptake value)를 이용하여 보다 정확하고 편리한 호흡동조 방사선치료의 metabolic target volume (MTV) 적용에 대한 유용성을 평가하고자 하였다. 평가를 위해 4D 팬텀에 임의의 인공산물을 만들어 PET 영상을 획득하였으며, 최대 SUV를 기준으로 임의로 설정한 50%, 30%, 그리고 5%의 SUV에서의 VOIs (Volumes Of Interest)와 호흡동조 방사선치료를 위한 4D-CT를 통해 획득한 호흡위상백분율에서 설정한 GTV (Gross Target Volume)을 비교하였다. 4D-CT를 통해 얻은 총합 GTV와 PET 영상의 30% SUV로 얻은 VOI와의 비교는 50%의 SUV로 얻은 VOI의 비교 결과보다 종(Longitudinal) 방향에서의 오차가 상당히 감소되었으며 4D 총합 CTV와 가장 일치하는 PET 영상은 5% SUV로 얻은 VOI로 관찰되었다. 4D PET/CT에서 전체 호흡의 25% 흡기에서 25% 호기까지 호흡위상백분율 영상의 30% SUV로 얻은 VOI는 IGRT (Image-guided radiation therapy)에 적용되는 4D-CT의 동일한 호흡위상백분율 영상에서 설정한 GTV와 비교한 결과, 최대 0.5 cm 이하로 잘 일치하였으며 4D PET의 5% SUV로 얻은 VOI의 경우 모든 방향에서 잘 일치하였다. 따라서 IGRT의 MTV 적용에 있어서 일반 PET 영상의 이용보다 4D PET 영상의 적용이 더 유용함을 보였다. 본 연구결과 현재 핵의학과에서 인체종양의 VOI를 30% SUV로 권고하고 있지만 30% 이하의 주변 SUV와 구분되는 최소 SUV를 선택해 적용한다면, 더욱 유용한 MTV 적용이 될 것으로 판단된다.

• 대한의용생체공학회:학술대회논문집
• /
• 대한의용생체공학회 1992년도 춘계학술대회
• /
• pp.58-63
• /
• 1992

The discrete cosine transform (DCT)-based progressive coding standard proposed by the International Standardization Orgnnization (ISO) Joint Photographic Experts Groups (JPEG) is investigated for medical image compression and the performance of the JPEG is compared to that of the full-frame bit-allocation (FFBA) technique. From the comparison, the JPEG standard appeals superior to the FFBA technique in the following aspects: 1) JPEG achieves higher compression than the FFBA technique with less mean square error. 2) Less Gibb's artifact is observed in the compressed images by the JPEG. 3) Computational time for the JPEG is about one-fourth or the FFBA technique. Other attractive points of the JPEG include: Implementation of the JPEG with VLSIs is commercially available in relative low price and the JPEG compression format can easily be interchangeable with other applications.