• 한국정보과학회논문지:소프트웨어및응용
• /
• 제37권2호
• /
• pp.110-119
• /
• 2010

본 논문에서는 MR 영상에서 밝기값 분포와 기울기 정보를 이용한 전립선 자동 분할 기법을 제안한다. 첫째, 적응적 밝기값 프로파일과 다해상도 기법을 이용하는 활성형상모델을 통해 전립선 표면을 추출한다. 둘째, 표면 형상의 지역적 최적화로 인한 흘을 방지하기 위하여 기하학 정보를 이용한 흘 제거 기법을 수행한다. 셋째, 해부학적으로 변이가 큰 표면 형상은 2차원 기울기 정보를 이용하여 보정한다. 이때, 보정된 표면 형상은 한정된 정점의 개수로 산정되어 매끄럽게 표현되지 않기 때문에 표면재구성 및 평활화 기법을 이용하여 부드러운 형상으로 표현한다. 제안방법의 평가를 위하여 육안평가와 정확성 평가 그리고 수행시간을 측정하였다. 정확성 평가는 두 명의 임상전문의의 수동분할 결과와 자동분할 결과 간의 평균거리차이와 중복볼륨비율을 측정하였다. 실험 결과 평균거리차이는 0.3${\pm}$0.21mm 측정되었고, 중복볼륨 비율은 96.31${\pm}$2.71% 측정되었다. 20명의 환자 데이터에 대한 전체 수행시간은 평균 16초로 측정되었다.

• 한국컴퓨터그래픽스학회논문지
• /
• 제29권5호
• /
• pp.21-30
• /
• 2023

부분신장절제술 전 수술 계획을 세우기 위해서는 신장 종양의 위치, 형태 및 수술 시 안전 마진 파악이 중요하므로 신장 종양을 정확히 분할하는 것이 필요하다. 그러나 신장 종양은 환자마다 위치 및 크기가 다양하며 소장과 비장 같은 주변 장기와 형태와 밝기값이 유사하여 신장 종양을 분할하는 것에 어려움이 있다. 본 논문에서는 레이블이 있는 데이터와 없는 데이터를 함께 사용하는 준지도학습 방법 중 하나인 평균-교사모델을 활용하여 신장의 여러 위치에서 발생하는 작은 크기의 신장 종양을 분할하기 위해 신장 위치 정보를 가지는 신장 로컬 가이드 맵을 이용해 신장 종양에 집중하는 평균-교사 네트워크를 제안하고, 신장 종양의 크기에 따른 성능을 분석한다. 실험 결과, 제안 방법은 신장 주변에 존재하는 종양의 위치를 찾기 위해 신장 로컬 가이드 맵을 사용하여 신장의 국소 정보를 고려함으로써 75.24%의 F1-score를 보였다. 특히 분할이 어려운 작은 크기의 종양에 대한 과소분할을 개선하였으며 nnU-Net보다 적은 양의 레이블 데이터를 사용하여도 13.9% 높은 F1-score를 보였다.

• International Journal of CAD/CAM
• /
• 제9권1호
• /
• pp.121-128
• /
• 2010

The purpose of this study is to clarify the mechanical behavior of human lumbar vertebrae (L3/L4) with and without fusion bone under physiological axial compression. The author has developed the program code to build the patient specific three-dimensional geometric model from the computed tomography (CT) images. The developed three-dimensional model provides the necessary information to the physicians and surgeons to visually interact with the model and if needed, plan the way of surgery in advance. The processed data of the model is versatile and compatible with the commercial computer aided design (CAD), finite element analysis (FEA) software and rapid prototyping technology. The actual physical model is manufactured using rapid prototyping technique to confirm the executable competence of the processed data from the developed program code. The patient specific model of L3/L4 vertebrae is analyzed under compressive loading condition by the FEA approach. By varying the spacer position and fusion bone with and without pedicle instrumentation, simulations were carried out to find the increasing axial stiffness so as to ensure the success of fusion technique. The finding was helpful in positioning the fusion bone graft and to predict the mechanical stress and deformation of body organ indicating the critical section.

• 한국컴퓨터정보학회논문지
• /
• 제22권12호
• /
• pp.125-130
• /
• 2017

Screw insertion surgery is frequently required among surgical procedures. Especially, very careful attention should be paid to the insertion of screw in the operation of the cervical vertebra. Therefore, there is a need for a guide that allows the surgeon to reliably and promptly perform treatment by calculating the desired insertion angle and length for screw insertion. In this study, the center and direction of the pedicle were calculated through 3D modeling and 3D vector numerical analysis using the CT or MRI image of the patient for the safe operation of the guide, and based on this, After that, we will implement surgical guide based on this.

• Korean Journal of Radiology
• /
• 제21권5호
• /
• pp.511-525
• /
• 2020

Chest X-ray radiography and computed tomography, the two mainstay modalities in thoracic radiology, are under active investigation with deep learning technology, which has shown promising performance in various tasks, including detection, classification, segmentation, and image synthesis, outperforming conventional methods and suggesting its potential for clinical implementation. However, the implementation of deep learning in daily clinical practice is in its infancy and facing several challenges, such as its limited ability to explain the output results, uncertain benefits regarding patient outcomes, and incomplete integration in daily workflow. In this review article, we will introduce the potential clinical applications of deep learning technology in thoracic radiology and discuss several challenges for its implementation in daily clinical practice.

• 전자공학회논문지 IE
• /
• 제43권4호
• /
• pp.21-27
• /
• 2006

본 논문은 사람의 뇌에 대한 자기공명영상에서 백질과 회백질을 분리하고 각각의 체적을 산출하기 위한 것이다. 정상인과 비정상인의 대뇌 영상으로부터 백질, 회백질, 뇌척수액을 분리하고, 분리된 조직의 체적을 계산한다. 본 논문에서는 뇌 MR영상에서 체적을 산출하고 백질과 회백질을 산출하기 위한 새로운 방법을 제안한다. 그리고 각 구성 비율에 따라 표현되는 명암 값 분석을 통한 대뇌 자기공명영상으로부터 백질 및 회백질을 추출할 수 있는 판별값을 결정하는 방법을 개발하였다. 각 슬라이스에 추출된 화소의 수를 이용하여 백질 및 회백질의 체적을 구하는 방법을 제안하였다. 그리고 환자의 뇌척수액/대뇌 체적비와 연령을 입력으로 받아 판별식을 통해 판별값을 계산하며, 계산된 판별값을 이용해 기준점과 비교함으로써 정상과 비정상을 진단하였다. 결과적으로 연령이 증가 할수록 백질과 회백질의 체적은 감소하고 뇌척수액의 체적은 증가하고 있는 것을 알 수 있었다.

• Imaging Science in Dentistry
• /
• 제43권3호
• /
• pp.163-169
• /
• 2013

Purpose: Tonsilloliths are calcifications found in the crypts of the palatal tonsils and can be detected on routine panoramic examinations. This study was performed to highlight the benefits of cone-beam computed tomography (CBCT) in the diagnosis of tonsilloliths appearing bilaterally on panoramic radiographs. Materials and Methods: The sample group consisted of 7 patients who had bilateral radiopaque lesions at the area of the ascending ramus on panoramic radiographs. CBCT images for every patient were obtained from both sides of the jaw to determine the exact locations of the lesions and to rule out other calcifications. The calcifications were evaluated on the CBCT images using Ez3D2009 software. Additionally, the obtained images in DICOM format were transferred to ITK SNAP 2.4.0 pc software for semiautomatic segmentation. Segmentation was performed using contrast differences between the soft tissues and calcifications on grayscale images, and the volume in mm3 of the segmented three dimensional models were obtained. Results: CBCT scans revealed that what appeared on panoramic radiographs as bilateral images were in fact unilateral lesions in 2 cases. The total volume of the calcifications ranged from 7.92 to $302.5mm^3$. The patients with bilaterally multiple and large calcifications were found to be symptomatic. Conclusion: The cases provided the evidence that tonsilloliths should be considered in the differential diagnosis of radiopaque masses involving the mandibular ramus, and they highlight the need for a CBCT scan to differentiate pseudo- or ghost images from true bilateral pathologies.

• 한국의학물리학회지:의학물리
• /
• 제31권3호
• /
• pp.111-123
• /
• 2020

Deep learning (DL) is a subset of machine learning and artificial intelligence that has a deep neural network with a structure similar to the human neural system and has been trained using big data. DL narrows the gap between data acquisition and meaningful interpretation without explicit programming. It has so far outperformed most classification and regression methods and can automatically learn data representations for specific tasks. The application areas of DL in radiation oncology include classification, semantic segmentation, object detection, image translation and generation, and image captioning. This article tries to understand what is the potential role of DL and what can be more achieved by utilizing it in radiation oncology. With the advances in DL, various studies contributing to the development of radiation oncology were investigated comprehensively. In this article, the radiation treatment process was divided into six consecutive stages as follows: patient assessment, simulation, target and organs-at-risk segmentation, treatment planning, quality assurance, and beam delivery in terms of workflow. Studies using DL were classified and organized according to each radiation treatment process. State-of-the-art studies were identified, and the clinical utilities of those researches were examined. The DL model could provide faster and more accurate solutions to problems faced by oncologists. While the effect of a data-driven approach on improving the quality of care for cancer patients is evidently clear, implementing these methods will require cultural changes at both the professional and institutional levels. We believe this paper will serve as a guide for both clinicians and medical physicists on issues that need to be addressed in time.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권3호
• /
• pp.958-979
• /
• 2023

Due to laws, regulations, privacy, etc., between 70-90 percent of providers do not share medical data, forming a "data island". It is essential to collaborate across multiple institutions without sharing patient data. Most existing methods adopt distributed learning and centralized federal architecture to solve this problem, but there are problems of resource heterogeneity and data heterogeneity in the practical application process. This paper proposes a collaborative deep learning modelling method based on the blockchain network. The training process uses encryption parameters to replace the original remote source data transmission to protect privacy. Hyperledger Fabric blockchain is adopted to realize that the parties are not restricted by the third-party authoritative verification end. To a certain extent, the distrust and single point of failure caused by the centralized system are avoided. The aggregation algorithm uses the FedProx algorithm to solve the problem of device heterogeneity and data heterogeneity. The experiments show that the maximum improvement of segmentation accuracy in the collaborative training mode proposed in this paper is 11.179% compared to local training. In the sequential training mode, the average accuracy improvement is greater than 7%. In the parallel training mode, the average accuracy improvement is greater than 8%. The experimental results show that the model proposed in this paper can solve the current problem of centralized modelling of multicenter data. In particular, it provides ideas to solve privacy protection and break "data silos", and protects all data.

• 산업경영시스템학회지
• /
• 제46권3호
• /
• pp.1-6
• /
• 2023

In the realm of dental prosthesis fabrication, obtaining accurate impressions has historically been a challenging and inefficient process, often hindered by hygiene concerns and patient discomfort. Addressing these limitations, Company D recently introduced a cutting-edge solution by harnessing the potential of intraoral scan images to create 3D dental models. However, the complexity of these scan images, encompassing not only teeth and gums but also the palate, tongue, and other structures, posed a new set of challenges. In response, we propose a sophisticated real-time image segmentation algorithm that selectively extracts pertinent data, specifically focusing on teeth and gums, from oral scan images obtained through Company D's oral scanner for 3D model generation. A key challenge we tackled was the detection of the intricate molar regions, common in dental imaging, which we effectively addressed through intelligent data augmentation for enhanced training. By placing significant emphasis on both accuracy and speed, critical factors for real-time intraoral scanning, our proposed algorithm demonstrated exceptional performance, boasting an impressive accuracy rate of 0.91 and an unrivaled FPS of 92.4. Compared to existing algorithms, our solution exhibited superior outcomes when integrated into Company D's oral scanner. This algorithm is scheduled for deployment and commercialization within Company D's intraoral scanner.