A rough set based data model for breast cancer mammographic mass diagnostics

Int. J. Biomedical Engineering and Technology, Vol. 18, No. 4, 2015 359

Copyright © 2015 Inderscience Enterprises Ltd.

A rough set based data model for breast cancer mammographic mass diagnostics

Aaron Don M. Africa* and Melvin K. Cabatuan Electronics and Communications Engineering Department, De La Salle University, 2401 Taft Avenue Manila, Philippines Email: [email protected] Email: [email protected] *Corresponding author

Abstract: Breast cancer is the principal cause of cancer deaths among women, and early diagnosis is critical to its survival. Mammography is the recommended diagnostic procedure for ages 40 years and older. However, the low precision rate of mammographic result leads to needless biopsies. Thus, in this paper, we present the application of rough set theory in the development of a data model to aid in physician’s recommendation for biopsy. In particular, we will utilise the data obtained at the Institute of Radiology of the University Erlangen-Nuremberg between 2003 and 2006. The results showed that the rough set approach successfully reduced the dimensionality of the aforementioned data set by approximately 47%, and the outcome rules were validated using empirical testing at 100%.

Keywords: rough set theory; breast cancer; biomedical engineering; decision support systems.

Reference to this paper should be made as follows: Africa, A.D.M. and Cabatuan, M.K. (2015) ‘A rough set based data model for breast cancer mammographic mass diagnostics’, Int. J. Biomedical Engineering and Technology, Vol. 18, No. 4, pp.359–369.

Biographical notes: Aaron Don M. Africa earned his BS degree in Electronics and Communications Engineering (ECE) from the University of Santo Tomas. He completed his Masters degree in ECE at the De La Salle University Manila. Presently, he is an Associate Professor of De La Salle University Manila and a PhD graduate in ECE of the said school.

Melvin K. Cabatuan received the BSc degree in Electronics and Communications Engineering from Cebu Institute of Technology University, Cebu, Philippines, in 2004; and MS degree in Engineering from NAIST located at Ikoma, Nara, Japan, in 2010. He joined the Electronics & Communications Engineering department of De La Salle University in 2011, where he is currently an Assistant Professor. His research interest involves machine learning applications to health informatics.

360 A.D.M. Africa and M.K. Cabatuan

1 Introduction

Progress in the control of breast cancer will require sustained and increased efforts to provide high-quality screening, diagnosis, and treatment to all segments of the population (DeSantis et al., 2013). Breast cancer or the growth of malignant cells in the breast is the leading cause of cancer mortality among women worldwide. In 2010, breast cancer killed around 425,000 women, and this trend has been increasing each year since 1980s (Forouzanfar et al., 2010). In fact, it is the most prevalent cancer among women in 145 countries (International Agency for Research on Cancer, 2012) as shown in Figure 1. However, it has been known that early detection of the disease coupled with proper treatment increases the rate of survival. Thus, appropriate breast cancer screening and diagnostics methods which are both effective and accessible to the general population are necessary to solve this global problem.

Figure 1 Distribution of breast cancer among women worldwide

Source: Cancer Research UK, World Cancer Factsheet 2012

Mammography or the application of low-energy X-rays to the breast tissue is among the most common breast cancer screening methods used today. Although proven to reduce breast cancer mortality on women ages 39–49 (Magnus et al., 2011), the role of mammography remains controversial due to high false positive rates leading to unnecessary biopsies (Goldman, 2012). This leads to serious problems such as psychological effects on women who are wrongly diagnosed (Bond et al., 2013).

There are several existing studies that address integration of technology to improve healthcare and/or particularly breast cancer screening and diagnostics. This includes research on telemedicine, technology impact on healthcare, breast cancer screening intervention management, and computer-assisted detection and diagnostic systems or investigations on its essential components. Byrd and Byrd (2013) claimed that

A rough set based data model for breast cancer 361

information technology generally affects the quality of healthcare in hospitals. Puentes et al. (2007) have examined the significant interplay between current technologies, i.e. wireless broadband, non-invasive sensors, emerging multimedia standards, and open source software, with telemedicine development. Kavitha and Chellamuthu (2013) have proposed a web service oriented medical healthcare application for the cancer disease prediction, based on pattern matching and Support Vector Machine (SVM) classifier with majority voting-based combination framework. Some studies focused on intervention management acknowledging the vital role of primary care and increasing breast cancer screening attendance (Baskaran et al., 2009; Baskaran et al., 2012). Furthermore, other studies have investigated computer-assisted breast cancer detection and diagnostic systems. Doi (2007), Rangayyan et al. (2007) and Tang et al. (2009) provided a comprehensive review of techniques for computer-aided detection and diagnosis of breast cancer, i.e. detection of calcifications, masses, architectural distortion, and bilateral asymmetry, including image enhancement and image retrieval. Karthik et al. (2014) presented the concept of a ‘virtual doctor’, an artificial medical diagnostic system based on hard (ex. images) and soft (ex. fever, headache, cough, and other symptoms) inputs. Similarly, Srivastava et al. (2013) have presented the design, analysis, and classifier evaluation for a computer aided diagnostics (CAD) tool for early breast cancer detection from mammograms; Janghel et al. (2012) have considered hybrid approaches for classification of breast cancer based on ensemble of Artificial Neural Network (ANN) and Evolutionary Neural Network (ENN) optimisation; and Kala et al. (2011) have constructed a mixture of complex modular neural network experts model for medical diagnosis. Additionally, Saraswathi and Srinivasan (2014) have developed a classification technique for mammogram images using Fully Complex-Valued Relaxation Networks (FCRN) based ensemble approach. Lastly, other studies have focused on essential components of breast cancer CAD systems like image enhancement, segmentation, data clustering, classification, and optimisation. Sriramkumar et al. (2013) have developed an intelligent system for the identification of micro-calcification clusters (segmentation) in digitised mammograms which can be used to identify tumours as benign or malignant while Sundaram et al. (2012) have presented a contrast enhancement of mammograms based on Histogram-Modified Local Contrast Enhancement (HMLCE). Selvi and Malmathanraj (2011) implemented Q learning algorithm for multilevel thresholding technique in mammographic image segmentation. Cherni et al. (2013) have examined breast cancer cells classification utilising computational intelligence techniques, i.e. fuzzy c-means (FCM), ANN, and genetic algorithm (GA). Chawla and Duhan (2014) have reviewed some of the applications of three new algorithms, i.e. biogeography-based optimisation, cuckoo search and bat algorithm, in various domains of biomedical engineering, i.e. clinical diagnosis, etc. Lai and Garibaldi (2013) examined two variations of semi-supervised Fuzzy c-Means (ssFCM) algorithms for Nottingham Tenovus Breast Cancer (NTBC) data set into the six clinically useful subgroups.

In this study, we present a rough set based data model for breast cancer mammographic mass diagnostics. This data model will reduce the decision rules necessary in the evaluation of malignant or benign mammograms and help physicians in their decision process for biopsy recommendations. This model may also be utilised as an input pre-processing for breast cancer CAD systems. The rules that are produced using rough set theory will then be validated using empirical testing. An implication for theory


and practice for this research is that diagnosing Breast cancer mammographic mass will be easier because fewer information on the symptoms are needed to determine the possible cause.

2 Mammographic mass data set

The mammographic mass database utilised in this research is obtained from Institute of Radiology, University of Erlangen-Nuremberg (Elter et al., 2007). The data set is composed of six attributes – BI-RADS assessment, age, shape, margin, density, and severity. The Breast Imaging Reporting and Data System (BI-RADS) formulated by American College of Radiology is a standard tool in reporting mammograms. It has seven different categories for mammographic lesions according to their likelihood of malignancy. These include the following ordinal designations with their corresponding recommendations: 0 – incomplete (additional imaging needed); 1 – negative (normal interval follow-up); 2 – benign (normal interval follow-up); 3 – probably benign (short-interval follow-up); 4 – suspicious abnormality (biopsy); 5 – highly suggestive of malignancy (appropriate action); and 6 – histological proven malignancy (appropriate therapy) (Obenauer et al., 2005). The age data represents the patient’s age in years ranging from around 20 to 90 years old. The shape describes the mammographic mass with the following nominal designation: 1 – round; 2 – oval; 3 – lobular; and 4 – irregular. The margin of mass is described as: 1 – circumscribed, 2 – microlobulated; 3 – obscured; 4 – ill-defined; and 5 – speculated. The mass density is represented by: 1 – high; 2 – equal (iso); 3 – low; and 4 – fat-containing. Finally, the goal field which is the severity is represented as 0 for benign and 1 for a malignant mass. The class distribution of the database contains a total of 104 instances: 56 benign and 48 malignant cases. The aforementioned attributes of the mammographic mass data sets are summarised in Table 1.

3 Overview

Rough Set Theory was developed by Palwak in 1982. It was based on the assumption that data can be associated with every object that belongs to it (Palwak, 1982). The Rough Set Theory was an accepted mathematical framework and is used in applications like image processing, neural computing and data mining. Rough Set Theory is used in data that is derived from experience. In this research, Rough Set Theory was applied in mammographic mass diagnostics. The data that was used is a mammographic mass data set derived from the experience of doctors with patients who consulted them for breast cancer checkup. Rough Set Data Explorer (ROSE) was the software used to implement the Rough Set Theory Algorithm in the mammographic mass data set. This program was developed at the Laboratory of Intelligent Decision Support Systems of the Institute of Computing Science in Poznan.

The rules are tested using empirical testing (Preece, 2001). Empirical testing is done by making each rule produced by Rough Set Theory checked with the information system to see if their values matched. The number of rules where the values of the symptoms matched the information system was counted and assigned to variable b. The total number of rules is assigned to variable a. The percent validity c was computed using this formula: c = (b/a) 100.


Table 1 Mammographic mass data

Attributes Values Type

1. BI-RADS Assessment

0 (incomplete)

Ordinal non-predictive

1 (negative)

2 (benign)

3 (probably benign)

4 (suspicious abnormality)

5 (high probability of malignancy)

6 (proven malignancy)

2. Age 18–96 (in years) Integer predictive

3. Shape

1 (round)

Nominal predictive 2 (oval)

3 (lobular)

4 (irregular)

4. Margin

1 (circumscribed)

Nominal predictive

2 (microlobulated)

3 (obscured)

4 (ill-defined)

5 (spiculated)

5. Density

1 (high)

Ordinal predictive 2 (equal/iso)

3 (low)

4 (fat-containing)

6. Severity 0 (benign)

Binominal target class 1 (malignant)

4 Data and results

4.1 Information system of data

The following are the data in the diagnosis of mammographic mass. The data was taken from Institute of Radiology, University of Erlangen-Nuremberg. The first 104 entries of the datasets are processed in this paper. Rough Set Data Explorer (ROSE) was used to implement the Rough Set Theory algorithm in the datasets (ROSE, 1999).

The attributes are:

A. BIRADS B. Age C. Shape

D. Margin E. Density F. Severity


The symptoms are BIRADS, Age, Shape, Margin and Density. The possible cause is severity. This attribute will have a value of 1 if the mammographic mass is present and 0 if it is not based on the dataset. A symptom will have a value of 2 if it is unknown. The summary of the data taken from Institute of Radiology, University of Erlangen-Nuremberg (Elter et al., 2007) is shown in Table 2.

Table 2 Information system of mammographic mass data

# A B C D E F # A B C D E F

1. 5 67 3 5 3 1 33. 5 45 4 5 3 1

2. 4 43 1 1 X 1 34. 5 55 4 4 3 0

3. 5 58 4 5 3 1 35. 4 46 1 5 2 0

4. 4 28 1 1 3 0 36. 5 54 4 4 3 1

5. 5 67 3 5 3 1 37. 5 57 4 4 3 1

6. 4 43 1 1 X 1 38. 4 39 1 1 2 0

7. 5 58 4 5 3 1 40. 4 81 1 1 3 0

8. 4 28 1 1 3 0 41. 4 77 3 X X 0

9. 5 74 1 5 X 1 42. 4 60 2 1 3 0

10. 4 65 1 X 3 0 43. 5 67 3 4 2 1

11. 4 70 X X 3 0 44. 4 48 4 5 X 1

12. 5 42 1 X 3 0 45. 4 55 3 4 2 0

13. 5 57 1 5 3 1 46. 4 59 2 1 X 0

14. 5 60 X 5 1 1 47. 4 78 1 1 1 0

15. 5 76 1 4 3 1 48. 4 50 1 1 3 0

16. 3 42 2 1 3 1 49. 4 61 2 1 X 0

17. 4 64 1 X 3 0 50. 5 62 3 5 2 1

18. 4 36 3 1 2 0 51. 5 44 2 4 X 1

19. 4 60 2 1 2 0 52. 5 64 4 5 3 1

20. 4 54 1 1 3 0 53. 4 23 1 1 X 0

21. 3 52 3 4 3 0 54. 2 42 X X 4 0

22. 4 59 2 1 3 1 55. 5 67 4 5 3 1

23. 4 54 1 1 3 1 56. 4 74 2 1 2 0

24. 4 40 1 X X 0 57. 5 80 3 5 3 1

25. X 66 X X 1 1 58. 4 23 1 1 X 0

26. 5 56 4 3 1 1 59. 4 63 2 1 X 0

27. 4 43 1 X X 0 60. 4 53 X 5 3 1

28. 5 42 4 4 3 1 61. 4 43 3 4 X 0

29. 4 59 2 4 3 1 62. 4 49 2 1 1 0

30. 5 75 4 5 3 1 63. 5 51 2 4 X 0

31. 2 66 1 1 X 0 64. 4 45 2 1 X 0

32. 5 63 3 X 3 0 65. 5 59 2 X X 1


Table 2 Information system of mammographic mass data (continued)

# A B C D E F # A B C D E F

66. 5 52 4 3 3 1 86. 4 67 4 5 3 0

67. 5 60 4 3 3 1 87. 5 44 4 4 2 1

68. 4 57 2 5 3 0 88. 3 68 1 1 3 1

69. 3 57 2 1 X 0 89. 4 57 X 4 1 0

70. 5 74 4 4 3 1 90. 5 51 4 X X 1

71. 4 25 2 1 X 0 91. 4 33 1 X X 0

72. 4 49 1 1 3 0 92. 5 58 4 4 3 1

73. 5 72 4 3 X 1 93. 5 36 1 X X 0

74. 4 45 2 1 3 0 94. 4 63 1 1 X 0

75. 4 64 2 1 3 0 95. 5 62 1 5 3 1

76. 4 73 2 1 2 0 96. 4 73 3 4 3 1

77. 5 68 4 3 3 1 97. 4 80 4 4 3 1

78. 5 52 4 5 3 0 98. 4 67 1 1 X 0

79. 5 66 4 4 3 1 99. 5 59 2 1 3 1

80. 5 70 X 4 X 1 100. 5 60 1 X 3 0 81. 4 25 1 1 3 0 101. 5 54 4 4 3 1 82. 5 74 1 1 2 1 102. 4 40 1 1 X 0 83. 4 64 1 1 3 0 103. 5 62 4 4 3 0 84. 5 60 4 3 2 1 104. 4 33 2 1 3 0 85. 5 67 2 4 1 0

4.2 Reduction of the information system

The information in Table 2 is inputted to the Rough Set Data Explorer (ROSE) in order to implement the Rough Set Theory on the mammographic mass dataset. A total of 55 rules were formulated. Table 3 shows the rough set rules of the mammographic mass data.

Table 3 Rough set rules of the mammographic mass data

Num. Rule Case Value of symptoms in cases

1 (C = 2) & (D = 1) & (E = X) => (F = 0) F=0 C = 2, D = 1, E = X

2 (B = 28) => (F = 0) F = 0 B = 28

3 (B = 23) => (F = 0) F = 0 B = 23

4 (B = 81) => (F = 0) F = 0 B = 81

5 (B = 50) => (F = 0) F = 0 B = 50

6 (B = 33) => (F = 0) F = 0 B = 33

7 (B = 40) => (F = 0) F = 0 B = 40

8 (A = 4) & (E = 2) => (F = 0) F = 0 A = 4, E = 2

9 (B = 45) & (C = 2) => (F = 0) F = 0 B = 45, C = 2


Table 3 Rough set rules of the mammographic mass data (continued)


10 (B = 49) => (F = 0) F = 0 B = 49

11 (A = 4) & (B = 64) => (F = 0) F = 0 A = 4, B = 64

12 (B = 63) => (F = 0) F = 0 B = 63

13 (A = 4) & (D = X) => (F = 0) F = 0 A = 4, D = X

14 (B = 55) => (F = 0) F = 0 B = 55

15 (D = X) & (E = 3) => (F = 0) F = 0 D = X, E = 3

16 (B = 78) => (F = 0) F = 0 B = 78

17 (A = 4) & (B = 57) => (F = 0) F = 0 A = 4, B = 57

18 (B = 52) & (D = 5) => (F = 0) F = 0 B = 52, D = 5

19 (B = 51) & (C = 2) => (F = 0) F = 0 B = 51, C = 2

20 (B = 25) => (F = 0) F = 0 B = 25

21 (B = 52) & (C = 3) => (F = 0) F = 0 B = 52, C = 3

22 (A = 2) => (F = 0) F = 0 A = 2

23 (A = 4) & (B = 67) => (F = 0) F = 0 A = 4, B = 67

24 (B = 62) & (D = 4) => (F = 0) F = 0 B = 62, D = 4

25 (B = 43) & (D = 4) => (F = 0) F = 0 B = 43, D = 4

26 (B = 60) & (C = 2) => (F = 0) F = 0 B = 60, C = 2

27 (B = 36) => (F = 0) F = 0 B = 36

28 (B = 67) & (E = 1) => (F = 0) F = 0 B = 67, E = 1

29 (B = 54) & (D = 4) => (F = 1) F = 1 B = 54, D = 4

30 (B = 58) => (F = 1) F = 1 B = 58

31 (B = 75) => (F = 1) F = 1 B = 75

32 (B = 66) & (D = 4) => (F = 1) F = 1 B = 66, D = 4

33 (D = 3) => (F = 1) F = 1 D = 3

34 (A = 5) & (B = 67) & (D = 5) => (F = 1) F = 1 A = 5, B = 67, D = 5

35 (B = 45) & (C = 4) => (F = 1) F = 1 B = 45, C = 4

36 (A = 5) & (C = 1) & (D = 5) => (F = 1) F = 1 A = 5, C = 1, D = 5

37 (B = 74) & (E = 3) => (F = 1) F = 1 B = 74, E = 3

38 (B = 42) & (C = 4) => (F = 1) F = 1 B = 42, C = 4

39 (B = 44) => (F = 1) F = 1 B = 44

40 (B = 80) => (F = 1) F = 1 B = 80

41 (B = 76) => (F = 1) F = 1 B = 76

42 (A = 5) & (E = 2) => (F = 1) F = 1 A = 5, E = 2

43 (B = 59) & (E = 3) => (F = 1) F = 1 B = 59, E = 3

44 (B = 51) & (D = X) => (F = 1) F = 1 B = 51, D = X

45 (A = 3) & (D = 1) & (E = 3) => (F = 1) F = 1 A = 3, D = 1, E = 3

46 (B = 43) & (D = 1) => (F = 1) F = 1 B = 43, D = 1


Table 3 Rough set rules of the mammographic mass data (continued)


47 (A = 5) & (C = X) => (F = 1) F = 1 A = 5, C = X

48 (B = 53) => (F = 1) F = 1 B = 53

49 (B = 48) => (F = 1) F = 1 B = 48

50 (B = 64) & (C = 4) => (F = 1) F = 1 B = 64, C = 4

51 (A = X) => (F = 1) F = 1 A = X

52 (B = 73) & (E = 3) => (F = 1) F = 1 B = 73, E = 3

53 (A = 5) & (B = 57) => (F = 1) F = 1 A = 5, B = 57

54 (B = 59) & (D = X) => (F = 1) F = 1 B = 59, D = X

55 (B = 54) & (C = 1) => (F = 0) OR (F = 1)

F = 0 OR

F = 1

B = 54, C = 1

4.3 Discussion of the results

The data set presented 104 numbers of rules. The rules were reduced to 55 numbers of rules giving a 47% percent decrease. Applying empirical testing to the rules, we find the value of a and b. From Table 3 the value of a is 55. By inspecting Table 2 and Table 3, the value of b is 55. The percent validity is computed using the F: c = (b/a) 100. Substituting a and b to the equation, the percent validity is computed as 100%. Empirical testing showed that the rules produced by the Rough Set Theory are valid. The result showed that the severity of breast cancer mammographic mass can be determined with only the minimum number of information using rough set theory. This research is useful because it can reduce the number of test needed for severity identification.

5 Conclusions and recommendations

This research presented a way to show the possible causes of breast cancer with only the essential symptoms. This is done by incorporating the Rough Set Theory in the datasets of mammographic mass. This method is useful because in diagnosing breast cancer there is often incomplete information to determine the possible cause. Applying the Rough Set Theory in the datasets, 100% validity was obtained using empirical testing. A limitation of this research is that the database must have sufficient data in order for the rough set algorithm to work.

A recommendation in this research is to apply a conflict resolution model in the rules that are determined. There are occasions when there will be conflicting rules and the system may have difficulty in which rule to fire. The next step of this research is to use it in other breast cancer datasets. This research only used mammographic mass data, in theory it can be used in other breast cancer datasets with a possible cause and symptoms.


References

Baskaran, V., Bali, R., Arochena, H., Naguib, R., Wheaton, M., Wallis, M. and Wickramasinghe, N. (2009) ‘A knowledge-based primary care approach to increase breast screening attendance’, International Journal of Biomedical Engineering and Technology, Vol. 2, No. 2, pp.172–188.

Baskaran, V., Guergachi, A., Shah, B., Sidani, S., Bali, R., Naguib, R. and Wickramasinghe, N. (2012) ‘Information technology-initiated interventions: a case study for the UK National Health Service Breast Screening Programme to improve screening attendance using a new intervention research framework’, International Journal of Biomedical Engineering and Technology, Vol. 8, No. 1, pp.1–17.

Bond, M., Payey, T., Welch, K., Cooper, C., Garside, R., Dean, S. and Hyde, C. (2013) ‘Systematic review of the psychological consequences of false-positive screening mammographs’, Health Technology Assessment, Vol. 17, No. 13.

Byrd, L. and Byrd, T. (2013) ‘Accuracy of information as a mediator of IT and quality of healthcare’, International Journal of Biomedical Engineering and Technology, Vol. 11, No. 3, pp.270–285.

Chawla, M. and Duhan, M. (2014) ‘Applications of recent metaheuristics optimisation algorithms in biomedical engineering: a review’, International Journal of Biomedical Engineering and Technology, Vol. 16, No. 3, pp.268–278.

Cherni, M., Sayadi, M. and Fnaiech, F. (2013) ‘AI tools in medical image analysis: efficacy of ANN for oestrogen receptor status assessment in immunohistochemical staining of breast cancer’, International Journal of Biomedical Engineering and Technology, Vol. 12, No. 1, pp.60–83.

DeSantis, C., Ma, J., Bryan, L. and Jemal, H. (2013) ‘Breast cancer statistics’, Cancer Journal for Clinicians, Vol. 64, No. 1.

Doi, K. (2007) ‘Computer-aided diagnosis in medical imaging: historical review, current status and future potential’, Computerized Medical Imaging and Graphics, Vol. 31, Nos. 4/5, pp.198–211.

Elter, M., Schulz-Wendlant, R., Welch. and Wittenberg, K. (2007) ‘The prediction of breast cancer biopsy outcomes using two CAD approaches that both emphasize an intelligible decision process’, Medical Physics, Vol. 34, No. 11, pp.4164–4172.

Forouzanfar, M., Foreman, K., Delossantos, A., Lozano, R., Lopez, A., Murray, M. and Naghavi, M. (2010) ‘Breast and cervical cancer in 187 countries between 1980 and 2010: a systematic analysis’, The Lancet, Vol. 378, No. 9801, pp.1461–1484.

Goldman, L. (2012) ‘Facility characteristics do not explain higher false positive rates in diagnostic mammography at facilities serving vulnerable women’, Medical Care.

International Agency for Research on Cancer (IARC) and Cancer Research UK (2012) World Cancer Factsheet, London, UK.

Janghel, R., Shukla, A. and Tiwari, R. (2012) ‘Hybrid computing based intelligent system for breast cancer diagnosis’, International Journal of Biomedical Engineering and Technology, Vol. 10, No. 1, pp.1–18.

Kala, R., Janghel, R., Tiwari, R. and Shukla, A. (2011) ‘Diagnosis of breast cancer by modular evolutionary neural networks’, International Journal of Biomedical Engineering and Technology, Vol. 7, No. 2, pp.194–211.

Karthik, R., Menaka, R., Kulkarni, S. and Deshpande, R. (2014) ‘Virtual doctor: an artificial medical diagnostic system based on hard and soft inputs’, International Journal of Biomedical Engineering and Technology, Vol. 16, No. 4, pp.329–342.

Kavitha, A. and Chellamuthu, C. (2013) ‘Online medical healthcare application for cancer disease prediction based on pattern matching and SVM classifier with MVC (PMSMVC) framework’, International Journal of Biomedical Engineering and Technology, Vol. 12, No. 2, pp.177–188.


Lai, D. and Garibaldi, J. (2013) ‘A preliminary study on automatic breast cancer data classification using semi-supervised fuzzy c-means’, International Journal of Biomedical Engineering and Technology, Vol. 13, No. 4, pp.303–322.

Magnus, M., Ping, M., Bourgeois J. and Magnus, J. (2011) ‘Effectiveness of mammographic screening in reducing breast cancer mortality in women aged 39-49 years: a meta analysis’, Journal of Women’s Health, Vol. 20, No. 6, pp.845–852.

Obenauer, S., Hermann, K. and Grabbe, E. (2005) ‘Applications and literature review of the BI-RADS classification’, European Radiology, Vol. 15, No. 5, pp.1027–1036.

Palwak, Z. (1982) ‘Rough Sets’, International Journal of Information in Computer Science, Vol. 11, No. 5, pp.341–356.

Preece, A. (2001) Evaluating Verification and Validation Methods in Knowledge Engineering. Available online at: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.17.3766 (accessed on 24 March 2009).

Puentes, J., Bali, R., Wickramasinghe, N. and Naguib R. (2007) ‘Telemedicine trends and challenges: a technology management perspective’, International Journal of Biomedical Engineering and Technology, Vol. 1, No. 1, pp.59–72.

Rangayyan, R., Ayres, F. and Desautels, J. (2007) ‘A review of computer-aided diagnosis of breast cancer: toward the detection of subtle signs’, Journal of the Franklin Institute, Vol. 344, No. 3, pp.312–348.

ROSE 2.0 (1999) Available online at: http://www.idss.cs.put.poznan.pl/rose.

Saraswathi, D. and Srinivasan, E. (2014) ‘An ensemble approach to diagnose breast cancer using fully complex-valued relaxation neural network classifier’, International Journal of Biomedical Engineering and Technology, Vol. 15, No. 3, pp.243–260.

Selvi, S. and Malmathanraj, R. (2011) ‘Mammogram tumour classification using Q learning’, International Journal of Biomedical Engineering and Technology, Vol. 7, No. 4, pp.339–352.

Sriramkumar, D., Malmathanraj, R., Mohan, R. and Umamaheswari, S. (2013) ‘Mammogram tumour classification using modified segmentation techniques’, International Journal of Biomedical Engineering and Technology, Vol. 13, No. 3, pp.218–239.

Srivastava, S., Sharma, N., Singh, S. and Srivastava, R. (2013) ‘Design, analysis and classifier evaluation for a CAD tool for breast cancer detection from digital mammograms’, International Journal of Biomedical Engineering and Technology, Vol. 13, No. 3, pp.270–300.

Sundaram, M., Ramar, K., Arumugam, N. and Prabin, G. (2012) ‘Histogram-modified local contrast enhancement for mammogram images’, International Journal of Biomedical Engineering and Technology, Vol. 9, No. 1, pp.60–71.

Tang, J., Rangayyan, R., Xu, J., Naqa, I. and Yang, Y. (2009) ‘Computer-aided detection and diagnosis of breast cancer with mammography: recent advances’, IEEE Transactions on Information Technology in Biomedicine, Vol. 13, No. 2, pp.236–251.

A rough set based data model for breast cancer mammographic mass diagnostics

Documents