명시야 현미경 영상에서의 세포 분할을 위한 이중 사전 학습 기법

이규현(Gyuhyun Lee),트란민콴(Tran Minh Quan),정원기(Won-Ki Jeong) 한국컴퓨터그래픽스학회 2016 한국컴퓨터그래픽스학회 학술대회 Vol.2016 No.7

본 논문은 명시야 (bright-field) 현미경 영상를 위한 데이터 기반 세포 분할 알고리즘을 제시한다. 제시된 알고리즘은 일반적인 사전 학습 기법과 다르게 동시에 두 개의 사전과 관련된 희소 코드 (sparse code)를 통해 정의된 에너지 함수의 최소화를 진행하게 된다. 두 개의 사전 중 하나는 명시야 영상에 대해 학습된 사전이고 다른 하나는 사람에 의해 수작업으로 세포 분할된 영상에 대해 학습된 것이다. 학습된 두 개의 사전을 세포 분할 될 새로운 입력 영상에 대해 적용하여 이와 관련된 희소코드를 획득한 후 픽셀 단위의 분할을 진행하게 된다. 효과적인 에너지 최소화를 위해 합성곱 희소코드(Convolutional Sparse Coding)와 Alternating Direction of Multiplier Method(ADMM)이 사용되었고 GPU를 사용하여 빠른 분산연산이 가능하다. 본 연구는 이전에 사용된 가변형 모델 (deformable model)을 이용한 세포 분할 방식과는 다르게 제시된 알고리즘은 세포 분할을 위해 사전 지식이 필요없이 데이터 기반의 학습을 통해서 쉽고 효율적으로 세포 분할을 진행할 수 있다. Cell segmentation is an important but time-consuming and laborious task in biological image analysis. An automated, robust, and fast method is required to overcome such burdensome processes. These needs are, however, challenging due to various cell shapes, intensity, and incomplete boundaries. A precise cell segmentation will allow to making a pathological diagnosis of tissue samples. A vast body of literature exists on cell segmentation in microscopy images [1]. The majority of existing work is based on input images and predefined feature models only – for example, using a deformable model to extract edge boundaries in the image. Only a handful of recent methods employ data-driven approaches, such as supervised learning. In this paper, we propose a novel data-driven cell segmentation algorithm for bright-field microscopy images. The proposed method minimizes an energy formula defined by two dictionaries – one is for input images and the other is for their manual segmentation results – and a common sparse code, which aims to find the pixel-level classification by deploying the learned dictionaries on new images. In contrast to deformable models, we do not need to know a prior knowledge of objects. We also employed convolutional sparse coding and Alternating Direction of Multiplier Method (ADMM) for fast dictionary learning and energy minimization. Unlike an existing method [1], our method trains both dictionaries concurrently, and is implemented using the GPU device for faster performance.

명시야 현미경 영상에서의 세포 분할을위한 이중 사전 학습 기법

이규현(Gyuhyun Lee),트란민콴(Tran Minh Quan),정원기(Won-Ki Jeong) 한국컴퓨터그래픽스학회 2016 컴퓨터그래픽스학회논문지 Vol.22 No.3

본 논문은 명시야 (bright-field) 현미경 영상를 위한 데이터 기반 세포 분할 알고리즘을 제시한다. 제시된 알고리즘은일반적인 사전 학습 기법과 다르게 동시에 두 개의 사전과 관련된 희소 코드 (sparse code)를 통해 정의된 에너지 함수의 최소화를 진행하게 된다. 두 개의 사전 중 하나는 명시야 영상에 대해학습된 사전이고 다른 하나는 사람에 의해 수작업으로 세포 분할된 영상에 대해학습된 것이다. 학습된두개의 사전을 세포 분할 될 새로운 입력 영상에 대해 적용하여 이와 관련된 희소코드를 획득한 후 픽셀단위의 분할을 진행하게 된다. 효과적인 에너지최소화를 위해합성곱희소코드(Convolutional Sparse Coding)와Alternating Direction of Multiplier Method(ADMM)이 사용되었고 GPU를 사용하여 빠른분산 연산이 가능하다. 본 연구는 이전에 사용된 가변형 모델 (deformable model)을이용한 세포 분할 방식과는다르게 제시된 알고리즘은 세포 분할을 위해 사전 지식이 필요없이 데이터 기반의 학습을 통해서 쉽고 효율적으로 세포 분할을 진행할 수 있다. Cell segmentation is an important but time-consuming and laborious task in biological image analysis. An automated, robust, and fast method is required to overcome such burdensome processes. These needs are, however, challenging due to various cell shapes, intensity, and incomplete boundaries. A precise cell segmentation will allow to making a pathological diagnosis of tissue samples. A vast body of literature exists on cell segmentation in microscopy images [1]. The majority of existing work is based on input images and predefined feature models only – for example, using a deformable model to extract edge boundaries in the image. Only a handful of recent methods employ data-driven approaches, such as supervised learning. In this paper, we propose a novel data-driven cell segmentation algorithm for bright-field microscopy images. The proposed method minimizes an energy formula defined by two dictionaries – one is for input images and the other is for their manual segmentation results – and a common sparse code, which aims to find the pixel-level classification by deploying the learned dictionaries on new images. In contrast to deformable models, we do not need to know a prior knowledge of objects. We also employed convolutional sparse coding and Alternating Direction of Multiplier Method (ADMM) for fast dictionary learning and energy minimization. Unlike an existing method [1], our method trains both dictionaries concurrently, and is implemented using the GPU device for faster performance.

t-SNE 기반 군집화 분석을 통한 교모세포종 환자의 항암 후보 약물의 반응성 측정

이성민(Sungmin Lee),최민석(Minsuk Choi),이장호(Jangho Lee ),이규현(GyuHyun Lee),정원기(Won-Ki Jeong),윤성로(Sungroh Yoon) 한국정보과학회 2017 한국정보과학회 학술발표논문집 Vol.2017 No.06

델타스폿 용접기를 이용한 Al5182 용접성 평가

박병선(Byungsun Park),이문용(Munyong Lee),이규현(Gyuhyun Lee) 한국자동차공학회 2013 한국자동차공학회 부문종합 학술대회 Vol.2013 No.5

Resistance spot welding has been widely used in the sheet metal joining processes because of its high productivity and convenience. Recently automobile industries are trying to replace partly aluminum alloy sheets. Among currently produced aluminum alloys, Al alloy sheet of Al-Mg (5000 series) are being tested. Despite the lower fusion temperature of aluminum, more energy is needed for aluminum welding than for welding steel. Delta spot welding machine used process tape. Process tape has many functions. Contaminant sticks to the tape, not to the electrode protection from coatings (e.g. anti-corrosion primer) and pick-up of base material (e.g. Al). In this paper, we suggest the lobe curve for Al5182 with 2.0t, 1.6t and 1.4t using delta spot welding machine. The lobe curves which could expressed weldablity. There were three combinations for same sheet which were Al5182 2.0t, 1.6t and 1.4t. The trend of low boundary and high boundary variation of lobe curve were analyzed with a viewpoint of the contact resistance and the heat input.

레이저 열처리를 이용한 보론강 표면 열처리 특성

박병선(Byungsun Park),이문용(Munyong Lee),이규현(Gyuhyun Lee) 한국자동차공학회 2011 한국자동차공학회 학술대회 및 전시회 Vol.2011 No.11

Many studies about the development of high strength and light weight body parts are in progress and improve the fuel efficiency. One of the methods that increase strength is Surface Hardening Treatment. Laser heat treatment is an effective technique used to improve stiffness of car bodies. An experimental investigation with diode laser system was carried out to study the effect of heat treatment on the boron steel. The table of the test term was drawn up by temperature of laser heat treatment, speed and cooling. The heat treatment characteristics of the laser beam are analyzed using hardness tests. The dimension in heat treatment part and heat treatment part zone was also measured through vickers hardness test.

UHSS급 열연 강판의 범퍼 빔 적용 기술

조국래(Kukrae Cho),박세호(Seho Park),이규현(Gyuhyun Lee),김성주(Seongju Kim) 한국자동차공학회 2013 한국자동차공학회 부문종합 학술대회 Vol.2013 No.5

유한 요소법을 이용한 고장력 강판 인서트 리벳 최적화

박병준(Byungjoon Park),이문용(Munyong Lee),이규현(Gyuhyun Lee),조해용(Haeyong Cho) 한국자동차공학회 2013 한국자동차공학회 부문종합 학술대회 Vol.2013 No.5

The use of light-weight material as aluminum has been being demanded gradually because the weight lightening of vehicles is excellent in fuel efficiency improvement and the driving performance improvement. In advanced automobile company overseas, Self Piercing Rivet technology that applies to the aluminum and Dissimilar materials, but there are little application cases in domestic market that are only rely on overseas technology skills. Therefore, I optimized design and made Prototype by Self Piercing Rivet technology with FEM Analysis through this research and verified the possibility of mass production for the Confidence Verification of Self Piercing Rivet technology by Strength and Fatigue Life and Corrosion Evaluation.

Mg 판재 AZ31 합금의 고온 인장 특성과 조직관찰

김래형(Raehyeong Kim),이문용(Munyong Lee),이규현(Gyuhyun Lee),김동욱(Dongok Kim) 한국자동차공학회 2011 한국자동차공학회 학술대회 및 전시회 Vol.2011 No.11

Magnesium alloy is currently expected to be widely used for high lightweight and high efficient materials in the automobile and electronics industry. Although Magnesium is excellent materials, it can’t be formed in the room temperature due to HCP structure. However in order to improve the magnesium formability, it is required by the investigation of forming at warm temperature range. This study aims to confirm characteristic of formability of AZ31 magnesium alloy sheet by the warm tensile test and microstructure analysis. The formability of AZ31 magnesium was evaluated at various temperatures, strain rate, strain rate sensitivity and thickness. Furthermore the result of microstructure of AZ31 magnesium shows how grain growth is changed by thickness.

마그네슘 AZ31재를 이용한 차체부품 개발

윤태욱(Taewook Yoon),김래형(Raehyug Kim),이문용(Moonyong Lee),이규현(Gyuhyun Lee) 한국자동차공학회 2012 한국자동차공학회 지부 학술대회 논문집 Vol.2012 No.11

This study indicated that development of car body part in using Mg AZ31. Although Mg alloy (AZ31) which consist of HCP structure is hard to form in room temperature, it can be formed in warm temperature more than 200℃. The reason why it can improve forming ability in warm temperature is that non basal slip is activated. Car body part is made up warm press forming die with system of improved non basal slip. Before development of warm press forming system, warm forming die is designed and made up FLD TEST with FLC curve, warm forming analysis with the decrease rate of the thickness. Throughout a various process parameters such as blank temperature in 250℃, forming speed 10mm/sec and BHF 1ton so on, the optimized process condition and the design specification result in. Throughout the result of this study, the development technology of car body part is considered.