A cooperative belief rule based decision support system for lymph node metastasis diagnosis in gastric cancer

doi:10.1016/j.knosys.2015.04.019

Knowledge-Based Systems

Volume 85, September 2015, Pages 62-70

https://doi.org/10.1016/j.knosys.2015.04.019 Get rights and content

Abstract

Lymph Node Metastasis (LNM) has become one of the most important prognostic factors regarding long-term survival in gastric cancer. As it is difficult for doctors to integrate multiple factors for a comprehensive analysis, Clinical Decision Support System (CDSS) is used to help the analysis. In this paper, a new Cooperative Belief Rule Based (CBRB) prototype CDSS is proposed. CBRB consists of two independent Belief Rule Base (BRB) systems and the final output is combined by the Evidential Reasoning (ER) approach. A corresponding new Cooperative CoEvolutionary Algorithm (CCEA) is proposed to train the proposed CBRB model that is nonlinear. A case study demonstrates that the proposed CDSS prototype can obtain the better performance than other CDSS.

Introduction

Gastric cancer has become one of the leading causes for cancer-related death around the world [1]. Lymph Node Metastasis (LNM) is one of most important prognostic factors regarding long-term survival [2], [3], [4], [5]. As such, diagnosing LNM accurately is very important. Currently, doctors diagnose LNM mainly according to the size of lymph node. However, large lymph nodes may be caused by inflammation, while small ones may be metastatic [6]. So a single lymph node size is not a strong predictor and more factors should be considered by applying Clinical Decision Support System (CDSS).

In previous research, several CDSSs have been proposed for LNM diagnosis, such as Artificial Neural Network (ANN) based CDSS [7], and Support Vector Machine (SVM) based CDSS [8]. However, both of them have some limitations. First of all, since they are black-box modeling methods, the reasoning process cannot be seen and doctors do not know which factors are more important for diagnosis. Moreover, as doctors play a critical role in diagnosis, it is important to diagnose LNM by using both clinical data and doctors’ knowledge [9].

According to the above analysis, a knowledge-based CDSS which can capture human judgments is more suitable for LNM diagnosis. In [9], a Bi-level Belief Rule Based (BBRB) CDSS was proposed by the authors. That CDSS is constructed based on a recently developed belief Rule-based Inference Methodology using the Evidential Reasoning approach (RIMER) [10]. The CDSS consists of two parts: (1) a Clinical domain knowledge base modeled by Belief Rule Base (BRB) and (2) a reasoning process supported by the Evidential Reasoning (ER) approach [11], [12], [13], [14], [15]. Compared to other knowledge based CDSSs, BBRB have some advantages as described in [9].

In BBRB, however, only the number and size of lymph nodes are utilized for LNM diagnosis. In fact, LNM is not only related with lymph node factors, but also with tumor factors [16], [17], [18], [19], [20]. To fully utilize the above factors, a Cooperative Belief Rule Based (CBRB) prototype CDSS is proposed in this paper. CBRB consists of two independent BRBs. One is used for modeling lymph node factors, while the other is used for modeling tumor factors. The final result is obtained by integrating the output of two BRBs, which is implemented by the ER approach. Meanwhile, as manually constructed belief rule may not be accurate, it is necessary to train CBRB. In this paper, a corresponding new Cooperative CoEvolutionary Algorithm (CCEA) is designed, which can be used to optimize the two BRBs and weight coefficients simultaneously.

The rest of the paper is organized as follows. The problem formulation is shown in Section 2. In Section 3, the CBRB CDSS prototype is presented. A new CCEA based method for optimizing CBRB is developed in Section 4. In Section 5, the proposed CDSS prototype for diagnosing LNM is presented. The validation of BBRB is discussed in Section 6. This paper is concluded in Section 7.

Section snippets

Problem formulation

Suppose that $T = [T_{1}, T_{2}, \dots, T_{M}]$ is the set of diagnostic factors which are extracted from tumor, while $L = [L_{1}, L_{2}, \dots, L_{N}]$ is the set of factors from lymph nodes, where M and N are the number of attributes for the two type of factors respectively. Suppose that $P^{T}$ and $P^{L}$ are the corresponding parameter vectors for the two BRBs, $ω = [ω_{1}, ω_{2}]$ is the set of weight coefficients that represent the relative importance for each BRB. In other words, the problem is in essence to construct a causal relationship

CBRB prototype CDSS

As CBRB is constructed on the basis of BRB, BRB will be briefly described at first. Then the proposed CBRB model will be presented.

New optimization model and CCEA for training CBRB

In this section, a new optimization model for training CBRB and a corresponding CCEA based optimization algorithm is proposed.

CBRB prototype CDSS for diagnosing LNM

In this section, the proposed CBRB CDSS prototype is utilized for LNM diagnosis. It consists of two parts: (1) constructing each BRB model in CBRB system and (2) the training process for optimizing CBRB.

An experimental case study

In this section, CBRB is validated using a set of real patient data. Just as in [9], the 2-cross-validation approach is used. To better analyze the performance, Confusion Matrix (CM) is utilized to show the results. In this matrix, each column represents the instances in a predicted stage, and each row represents the instances in an actual stage. In CM, two evaluated indexes are used, that is: (1) User Accuracy (UA); (2) Procedure Accuracy (PA). UA represents the ratio of the number of

Conclusions

In this paper, a Cooperative Belief Rule Based (CBRB) CDSS prototype was proposed for LNM diagnosis in gastric cancer. CBRB consists of two independent BRBs and the final output is obtained by using the ER approach. In addition, a corresponding CCEA based method was proposed for training CBRB. By utilizing both the tumor and lymph node features, CBRB can obtain better diagnostic performance. A comparative case study on current CDSSs showed that CBRB performed better than other methods, needless

Acknowledgments

The authors would like to thank their collaborators of Beijing Cancer Hospital for their suggestions and impressive work and thank Professor Dong-Ling Xu from Manchester Business School, The University of Manchester for her suggestive comments. This research was supported in part the National Basic Research Program (973 Program) of China (No. 2013CB329402), in part by the National Natural Science Foundation of China (Nos. 61173090, 61173092, 61303119, 61271302, 61272282), in part by the

References (36)

C.W. Wu
Nodal dissection for patients with gastric cancer: a randomized controlled trial
Lancet Uncoil.
(2006)
Zhi-Guo Zhou et al.
An evidential reasoning based model for diagnosis of lymph node metastasis in gastric cancer
BMC Med. Inform. Dec. Making
(2013)
Zhi-Guo Zhou et al.
A bi-level belief rule based decision support system for diagnosis of lymph node metastasis in gastric cancer
Knowl.-Based Syst.
(2013)
J.B. Yang
The evidential reasoning approach for MADA under both probabilistic and fuzzy uncertainties
Euro. J. Operat. Res.
(2006)
Y.M. Wang
Environmental impact assessment using the evidential reasoning approach
Euro. J. Operat. Res.
(2006)
Z.J. Zhou
A sequential learning algorithm for online constructing belief-rule-based systems
Exp. Syst. Appl.
(2010)
Z.J. Zhou
A model for real-time failure prognosis based on hidden Markov model and belief rule base
Euro. J. Operat. Res.
(2010)
J.B. Yang
Rule and utility based evidential reasoning approach for multi-attribute decision analysis under uncertainties
Euro. J. Operat. Res.
(2001)
D.M. Parkin
Global cancer statistics
Can. J. Clinic.
(2005)
D. Cunningham et al.
East meets west in the treatment of gastric cancer
New Eng. J. Med.
(2007)

K. Ohkuma et al.

Computed tomography for stage of gastric cancer

Jap. J. Can. Clin.

(1986)

D.D. Smith

Impact of total lymph node count on stage and survival after gastrectomy for gastric cancer: data from a large US-population database

J. Clin. Oncol.

(2005)

H. Elfriede

Artificial neural network for prediction of lymph node metastases in gastric cancer: a phase II diagnostic study

Ann. Surg. Oncol.

(2004)

X.P. Zhang

Support vector machine model for diagnosis of lymph node metastasis in gastric cancer with multi-detector computed tomography: a preliminary study

BMC Can.

(2011)

J.B. Yang

Belief rule-base inference methodology using the evidential reasoning approach–RIMER

IEEE Trans. Syst., Man, Cybernet.-Part A: Syst. Hum.

(2006)

J.B. Yang et al.

A general multi-level evaluation process for hybrid MADM with uncertainty

IEEE Trans. Syst., Man, Cybernet.-Part A: Syst. Hum.

(1994)

J.B. Yang et al.

An evidential reasoning approach for multiple attribute decision making with uncertainty

IEEE Trans. Syst., Man, Cybernet. – Part A: Syst. Hum.

(1994)

J.B. Yang et al.

On the evidential reasoning algorithm for multiple attribute decision analysis under uncertainty

IEEE Trans. Syst., Man, Cybernet. – Part A: Syst. Hum.

(2002)

Cited by (60)

An ensemble belief rule base model for pathologic complete response prediction in gastric cancer
2023, Expert Systems with Applications
It is well known that the decision-making on treating gastric cancer is usually the summary of several experts’ advice. Moreover, the interpretability and reliability of a model used to assist doctors with final decisions are very important in gastric cancer. Thus, this paper designed an ensemble belief rule base (EnBRB) model to ensemble multiple BRB models and predict the statement of pathological complete response (pCR), aiming to simulate the expert consultation widely used for clinical decisions on gastric cancer. In EnBRB, five BRB models were built individually using experts’ knowledge and trained through patient treatment information. Furthermore, the final output was calculated via the evidential reasoning (ER) based ensemble strategy on the results of each BRB for a more reliable decision. Then, to improve the model’s performance, a two-stage approach with a differential evolution (DE) algorithm was designed to enhance the performance of EnBRB. The experimental results demonstrated that a higher accuracy ( $0.9296 \pm 0.0521$ ) and AUC ( $0.9570 \pm 0.0368$ ) were obtained by our EnBRB. Moreover, the difference between sensitivity and specificity achieved by EnBRB was smaller than other comparison methods.
Extended belief rule base with ensemble imbalanced learning for lymph node metastasis diagnosis in endometrial carcinoma
2023, Engineering Applications of Artificial Intelligence
Lymph node metastasis (LNM) constitutes one of the main prognostic factors for long-term survival in endometrial carcinoma (EC). However, the previous studies on LNM diagnosis failed to consider both model interpretability and class imbalance. In this study, the extended belief rule base (EBRB) expert system is introduced to develop a novel EBRB-based LNM diagnosis model. First, the interpretability of the EBRB expert system is investigated to demonstrate the feasibility on LNM diagnosis; Second, imbalanced learning is introduced to improve rule generation scheme for constructing base EBRBs; Third, by considering the trust of base EBRBs and base diagnoses, ensemble learning is introduced to improve rule inference scheme for diagnosing final LNM. In the case study, real EC patient data collected from Fujian Provincial Maternity and Children's Hospital are used to verify the effectiveness of the proposed EBRB-based model by comparing with the variants of rule generation schemes and rule inference schemes, as well as some machine learning-based LNM diagnosis models. The comparative results showed that the proposed EBRB-based model has better sensitivity, specificity, and geometric mean in diagnosing LNM for EC patients.
A fast belief rule base generation and reduction method for classification problems
2023, International Journal of Approximate Reasoning
The belief rule-based (BRB) system has been one of the most significant rule-based systems for its ability to deal with various kinds of information under uncertainty, and it has shown great potential for classification problems. However, the combinatorial explosion problem hugely limits the application of the BRB system as excessive rules could not only increase the computation cost, but also impact the performance of the BRB system. Therefore, motivated by this problem, this paper proposed a fast and accurate belief rule base generation and reduction method. Firstly, a fast belief rule base generation method is introduced, where similar rules are grouped and combined to ensure all the possible situations are covered without generating excessive rules. Then, a redundancy-based belief rule base reduction method is proposed, where the redundancy degree of the belief rule that represents the degree to which a belief rule is affected by other rules is introduced, and it is calculated to identify redundant rules. Furthermore, the conventional evidential reasoning (ER)-based inference process is retained from conventional BRB systems. Thirty classification benchmarks from the well-known UCI machine learning repository are tested to validate the effectiveness of the proposed method, and the results are compared with other rule-based systems, improved BRB systems, and other machine learning methods. Comparison results show that the proposed method could effectively reduce the size of the BRB without costing its accuracy. Furthermore, sensitivity analysis and robustness analysis are conducted, which further show the effectiveness and robustness of the proposed method.
An interpretable method for inertial platform fault diagnosis based on combination belief rule base
2023, Measurement: Journal of the International Measurement Confederation
As a key component of navigation equipment, the inertial platform needs high reliability, so an accurate and interpretable method for fault diagnosis is necessary. The belief rule base (BRB) has strong interpretability and its ability to process semi-quantitative information. However, the high complexity of inertial platforms leads to the existence of multiple indicators for use in fault diagnosis, which leads to the combinatorial explosion of rules problem in BRB. Considering this problem, an interpretable method for inertial platform fault diagnosis based on combination BRB is proposed here. This paper provides two main contributions: 1) A BRB optimization method considering the correlation coefficient is proposed, which maintains the interpretability of sub-BRBs in the combination BRB, and 2) a method is proposed to calculate the sub-BRB weights based on the precision of the sub-BRBs and the occurrence times of repeated indicators, thus ensuring that the fusion process is explainable when the ER algorithm is used for the fusion of sub-BRBs. The proposed method ensures the interpretability of the final result during fault diagnosis while maintaining the reliability of the inertial platform. Finally, the effectiveness of the proposed method is verified against state-of-the-art methods using actual operation data from an inertial platform.
Belief rule-base expert system with multilayer tree structure for complex problems modeling
2023, Expert Systems with Applications
Belief rule-base (BRB) expert system is one of recognized and fast-growing approaches in the areas of complex problems modeling. However, the conventional BRB has to suffer from the combinatorial explosion problem since the number of rules in BRB expands exponentially with the number of attributes in complex problems, although many alternative techniques have been looked at with the purpose of downsizing BRB. Motivated by this challenge, in this paper, multilayer tree structure (MTS) is introduced for the first time to define hierarchical BRB, also known as MTS-BRB. MTS-BRB is able to overcome the combinatorial explosion problem of the conventional BRB. Thereafter, the additional modeling, inferencing, and learning procedures are proposed to create a self-organized MTS-BRB expert system. To demonstrate the development process and benefits of the MTS-BRB expert system, case studies including benchmark classification datasets and research and development (R&D) project risk assessment have been done. The comparative results showed that, in terms of modelling effectiveness and/or prediction accuracy, MTS-BRB expert system surpasses various existing, as well as traditional fuzzy system-related and machine learning-related methodologies.
Disjunctive belief rule-based reasoning for decision making with incomplete information
2023, Information Sciences
Witnessing the growth of inaccurate data with uncertainty and ambiguity, the belief rule base system is attracting increasing research interest because of its usage in decision-making problems as a disjunctive paradigm particularly when there is incomplete information. In this paper, we devise a novel method that employs fuzzy association rules to optimize disjunctive belief rule base (DBRB) systems regarding incomplete information in decision-making and achieve a significant performance improvement compared to the existing ones. First, we devise a method for constructing a fuzzy representation base from history data by using the triangular membership functions to blur history information. Then, a method is proposed to construct association rules based on the fuzzy representation base, in which the association rules are found regarding the missing input information by carrying out reasoning to complete the data and then making decisions on the data filled with DBRB system. Lastly, we evaluate the effectiveness of our proposed method against the public test data set, demonstrating that our method outperforms other baselines for decision-making with incomplete information.

View all citing articles on Scopus

View full text

A cooperative belief rule based decision support system for lymph node metastasis diagnosis in gastric cancer

Abstract

Introduction

Section snippets

Problem formulation

CBRB prototype CDSS

New optimization model and CCEA for training CBRB

CBRB prototype CDSS for diagnosing LNM

An experimental case study

Conclusions

Acknowledgments

Lancet Uncoil.

BMC Med. Inform. Dec. Making

Knowl.-Based Syst.

Euro. J. Operat. Res.

Euro. J. Operat. Res.

Exp. Syst. Appl.

Euro. J. Operat. Res.

Euro. J. Operat. Res.

Global cancer statistics

Can. J. Clinic.

East meets west in the treatment of gastric cancer

New Eng. J. Med.

Computed tomography for stage of gastric cancer

Jap. J. Can. Clin.

Impact of total lymph node count on stage and survival after gastrectomy for gastric cancer: data from a large US-population database

J. Clin. Oncol.

Artificial neural network for prediction of lymph node metastases in gastric cancer: a phase II diagnostic study

Ann. Surg. Oncol.

Support vector machine model for diagnosis of lymph node metastasis in gastric cancer with multi-detector computed tomography: a preliminary study

BMC Can.

Belief rule-base inference methodology using the evidential reasoning approach–RIMER

IEEE Trans. Syst., Man, Cybernet.-Part A: Syst. Hum.

A general multi-level evaluation process for hybrid MADM with uncertainty

IEEE Trans. Syst., Man, Cybernet.-Part A: Syst. Hum.

An evidential reasoning approach for multiple attribute decision making with uncertainty

IEEE Trans. Syst., Man, Cybernet. – Part A: Syst. Hum.

On the evidential reasoning algorithm for multiple attribute decision analysis under uncertainty

IEEE Trans. Syst., Man, Cybernet. – Part A: Syst. Hum.