간 경변 진단시 신경망을 이용한 분류기 구현 = Implementation of the Classification using Neural Network in Diagnosis of Liver Cirrhosis

다국어 초록 (Multilingual Abstract)

This paper presents the proposed a classifier of liver cirrhotic step using MR(magnetic resonance) imaging and hierarchical neural network. The data sets for classification of each stage, which were normal, 1type, 2type and 3type, were analysis in the number of data was 231. We extracted liver region and nodule region from T1-weight MR liver image. Then objective interpretation classifier of liver cirrhotic steps. Liver cirrhosis classifier implemented using hierarchical neural network which gray-level analysis and texture feature descriptors to distinguish normal liver and 3 types of liver cirrhosis. Then proposed Neural network classifier learned through error back-propagation algorithm. A classifying result shows that recognition rate of normal is 100%, 1type is 82.8%, 2type is 87.1%, 3type is 84.2%. The recognition ratio very high, when compared between the result of obtained quantified data to that of doctors decision data and neural network classifier value. If enough data is offered and other parameter is considered, this paper according to we expected that neural network as well as human experts and could be useful as clinical decision support tool for liver cirrhosis patients.

국문 초록 (Abstract)

자기공명영상과 계층적 신경망을 이용하여 간경변증을 단계별로 분류하고자 하였다. 내원한 231명의 데이터를 분석하였으며, 각 단계별 분류는 정상, 1, 2, 3단계로 분류하였다. T1강조 자기공명 간 영상으로부터 정상 간 실질과 간 경변 결절을 추출하고, 간 경화증의 단계를 객관적으로 해석 분류하였다. 간 경변 분류기 구현은 계층적 신경망을 이용하였고, 명암도 분석과 간 결절 특성을 통하여 정상간과 3단계의 간 경변으로 구분하였다. 제안한 신경망 분류기는 오류 역전파 알고리듬을 이용하였다. 분류결과 인식율이 정상군은 100%, 1 단계는 82.8%, 2 단계는 87.1%, 3 단계는 84.2%의 분류율을 나타내었다. 신경망 분류 결과와 전문의 판독 결과를 서로 비교한 결과 인식률은 매우 높게 나타났다. 만일 더욱더 충분한 데이터나 파라미터를 가지고 지속적으로 수행한다면 간 경변 환자들에게 임상적으로 지원하는 도구뿐만 아니라 의료전문 신경망으로도 기대된다.

참고문헌 (Reference)

• 1 "Ultrasonography in Gastroenterology" 1985.
• 2 "The morphology of cirrhosis. Journal of Clinical Pathology" 395-414, 1978
• 3 "Textureal features for Image Classification" and cybernetics : 610-621, vol.smc-3
• 4 "Texture Feature for Ultrasonic Liver Images" 11 : -2, 1992
• 5 "Segmentation of medical images" 11 : 486-503, 1995
• 6 "Neurons with graded response have collective computational properties like those of two-state neurons" 81 : 3088-3092, 1984
• 7 "Neural network design" PWS publishing company 1996.
• 8 "Morphological Segmentation of Textured Cell Images" 40 (40): 1996
• 9 "Image analysis using mathematical morphology" pami-9 : -4, 1987
• 10 "Histogram Equalization" 173-180,
• 11 "Estimation of the Scatterer Distribution of the Cirrhotic Liver using Ultrasonic Image" 1998
• 12 "Classification of Scattering Objects by Regular Arrys of Scatterers using Fourier Analysis" 346-360, 1983
• 13 "Cirrhosis and hepatoma in alcoholics" 77-85, 1966
• 14 "Bounded Gray-Level Morphology and Its Applications to Image Representation" 5 (5): 1996
• 15 "An improved seeded region growing algorithm" 18-, 1997
• 16 "An Automatic Diagnostic System for CT Liver Image Classification" 45 (45): 1998
• 17 "Algorithms for Image processing and Computer vision." 155-171, 1997
• 18 "Adenomatous hyperplasticnodules in the cirrhotic liver differentiation from hepatocellularcarcinoma with MR imaging" 123-126, 1989
• 19 "A simple neural network generating an interactive memory" 14 : 197-220, 1972
• 20 "A simple neural network generating an interactive memory" 14 : 197-220, 1972
• 21 "A Theoretical Comaprison of Texture Algorithms" 204-222, 1980