Novel structural descriptors for automated colon cancer detection and grading

被引:37
作者
Rathore, Saima [1 ,2 ]
Hussain, Mutawarra [1 ]
Iftikhar, Muhammad Aksam [1 ,3 ]
Jalil, Abdul [1 ]
机构
[1] Pakistan Inst Engn & Appl Sci, DCIS, Islamabad, Pakistan
[2] Univ Azad Jammu & Kashmir, DCS&IT, Muzaffarabad, Azad Kashmir, Pakistan
[3] Comsats Inst Informat Technol, Lahore, Pakistan
关键词
Colon cancer; Colon cancer detection; Colon cancer grading; Colon classification; AMINO-ACID-COMPOSITION; IMAGES; SEGMENTATION; PREDICTION;
D O I
10.1016/j.cmpb.2015.05.008
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
The histopathological examination of tissue specimens is necessary for the diagnosis and grading of colon cancer. However, the process is subjective and leads to significant inter/intra observer variation in diagnosis as it mainly relies on the visual assessment of histopathologists. Therefore, a reliable computer-aided technique, which can automatically classify normal and malignant colon samples, and determine grades of malignant samples, is required. In this paper, we propose a novel colon cancer diagnostic (CCD) system, which initially classifies colon biopsy images into normal and malignant classes, and then automatically determines the grades of colon cancer for malignant images. To this end, various novel structural descriptors, which mathematically model and quantify the variation among the structure of normal colon tissues and malignant tissues of various cancer grades, have been employed. Radial basis function (RBF) kernel of support vector machines (SVM) has been employed as classifier in order to classify/grade colon samples based on these descriptors. The proposed system has been tested on 92 malignant and 82 normal colon biopsy images. The classification performance has been measured in terms of various performance measures, and quite promising performance has been observed. Compared with previous techniques, the proposed system has demonstrated better cancer detection (classification accuracy = 95.40%) and grading (classification accuracy = 93.47%) capability. Therefore, the proposed CCD system can provide a reliable second opinion to the histopathologists. (C) 2015 Elsevier Ireland Ltd. All rights reserved.
引用
收藏
页码:92 / 108
页数:17
相关论文
共 27 条
[1]   Color Graphs for Automated Cancer Diagnosis and Grading [J].
Altunbay, Dogan ;
Cigir, Celal ;
Sokmensuer, Cenk ;
Gunduz-Demir, Cigdem .
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 2010, 57 (03) :665-674
[2]   MALIGNANT MESOTHELIOMA OF THE PLEURA - INTEROBSERVER VARIABILITY [J].
ANDRION, A ;
MAGNANI, C ;
BETTA, PG ;
DONNA, A ;
MOLLO, F ;
SCELSI, M ;
BERNARDI, P ;
BOTTA, M ;
TERRACINI, B .
JOURNAL OF CLINICAL PATHOLOGY, 1995, 48 (09) :856-860
[3]  
[Anonymous], 2012, CANC FACTS FIG
[4]  
[Anonymous], 2005, DATA MINING
[5]  
[Anonymous], 2008, A Practical Guide to Support Vector Classification
[6]   MEAN SHIFT, MODE SEEKING, AND CLUSTERING [J].
CHENG, YZ .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 1995, 17 (08) :790-799
[7]   Some remarks on protein attribute prediction and pseudo amino acid composition [J].
Chou, Kuo-Chen .
JOURNAL OF THEORETICAL BIOLOGY, 2011, 273 (01) :236-247
[8]  
Comaniciu D, 2000, PROC CVPR IEEE, P142, DOI 10.1109/CVPR.2000.854761
[9]   Robust analysis of feature spaces: Color image segmentation [J].
Comaniciu, D ;
Meer, P .
1997 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, PROCEEDINGS, 1997, :750-755
[10]  
Esgiar A N, 1998, IEEE Trans Inf Technol Biomed, V2, P197, DOI 10.1109/4233.735785