Diagnosis of gastrointestinal disorders using DIAGNET
Introduction
On analyzing recent development, it becomes clear that the trend is to develop new methods for computer decision making in medicine and to evaluate critically these methods in clinical practice. At present, medical expert systems help doctors to determine definitive diagnosis or a range of alternative diagnosis.
The conventional approach to build medical expert system requires the formulation of rules by which the input data can be analyzed. The formulation of such rules is very difficult with large sets of input data. In order to overcome the difficulty, artificial neural network (ANN) has been applied as an alternative to conventional rule-based expert system. ANNs can be trained without encapsulating the knowledge derived from these rules. Hence ANN has been found to be more helpful than a traditional medical expert system in the diagnosis of diseases. For example, patients may not have similar signs and symptoms when result is the same disease. The diseases of the patients cannot be classified into a single class unless some more measurements and tests are made to solve ambiguity.
Section snippets
Previous survey of medical applications involving Neural Network
Nowadays, multilayer perceptron appears to perform as an optimal generic modeling tool. It has been recognized in the medical literature that neural network has much contributed for modeling of cancer survival. Nevertheless, the interest highlighted by recent reviews of applications to critical care staging of prosthetic cancer and physical medicine and rehabilitation. Another study of particular interest addressed the automated cytodiagnosis of fine needle aspirates of the breast from 10
Combination of ANN and fuzzy logic
ANN models are inherently nonlinear and fault tolerant but do not facilitate to accept the real life inputs which is in linguistic relative terms such as low, normal, high and very high. The fuzzy logic is capable of modeling vagueness, handling uncertainty and supporting human type reasoning. Fuzzy relations are defined for the frequency of occurrence of symptoms with diseases and strength of confirmation of symptoms for diseases. So, the disadvantage of ANN can be overcome by adding the fuzzy
Backpropagation neural network (BPNN)
An artificial neural network is a computer program consisting of a simple processing unit, analogous to neurons. These processing units or nodes are interconnected by weights, analogous to synaptic connections in the brain. The weighted sum of all signals reaching a node is compared with a threshold. If the signal exceeds the threshold, the node fires. If not, the node remains quiescent. Memory is distributed throughout the network, and the complexity of the network is derived from the density
Radial basis functions neural network
The Radial Basis Function Neural Network (RBFNN) consists of three layers with extremely different roles. The input layer is made up of source nodes (sensory units) that connect the network to its environment. The second layer, only hidden layer in the network, applies a nonlinear transformation from the input layer to hidden layer. The output layer is linear, supplying the response of the network to the activation pattern (signal) applied to the input layer. The structure of RBFNN is shown in
System design
By oral interview with patients, the linguistic inputs are collected and recorded in the format given in the Appendix 1. At the front end of the model, an user interface offers to select symptoms and signs of gastrointestinal disorders. Based on the patients data, the front end user interface is filled. Symptoms and signs are either in crisp type or fuzzy nature. Crisp type of data are entered into enter two values either yes or no. The fuzzy nature of data has three attributes viz “mild”,
Data set
The data sets used in this study are collected from Raja Muthiah Medical College and Hospital, Annamalai University. This study is confined with gastrointestinal disorders. The most 15 common diseases related to gastrointestinal disorders are classified into the following four groups.
Group I: Hepatic diseases Amoebic liver abscess Biliary stricture Cholecystitis Cirrhosis of liver Ascites Carcinoma stomach Gastric erosion Gastric ulcer Peptic ulcer Oesophagitis Pelvic inflammatory disease Hodgkin’s lymphoma Appendicitis Appendicular
Group II: Gastric diseases
Group III: Urogenital infections
Group IV: Lympho reticular system diseases
Group V: Large intestine diseases
DIAGNET
DIAGNET is combination of backpropagation and radial basis functions neural network model. This new model is proposed for diagnosis of diseases in this research work. The structure of DIAGNET is given in Fig. 2. The symptoms and signs of gastrointestinal disorders are applied to both BPNN and RBFNN separately and their weight values are obtained during training. The BPNN with 115L 58N 15N provides the best performance where L denotes a linear unit and N denotes a nonlinear unit. The integer
Results and discussion
The following performance indicators are calculated: test efficiency (ratio of the number of correct diagnoses to the total number of patients), sensitivity (ratio of true positive diagnoses to true positive + false negative), specificity (ratio of true negative diagnoses to true negative + false positive), Positive Predictive Value (PPV, ratio of true positive to true positive + false positive) and Negative Predictive Value (NPV, is the ratio of the true negative to true negative + false negative).
Conclusion
When a medical expert works with large collection of data he should know the possibilities and dangers of computational methods in dealing with large stored information. On the other hand computer scientists working on medical software should be exposed to medical data analysis as well as specific purpose of medical knowledge. Computers in medicine cannot replace the medical expert in diagnosis or therapeutic decision making. However, computers are in general, and especially neural network and
Acknowledgements
We thank Dr. David Arumainayagam, Professor of General Medicine. RMMCH, Annamalai University and Dr. M. M. Kavitha, Tanjore Medical College, Tanjore for their help in carrying out this work effectively.
References (16)
Applications of neural networks to clinical medicine
Lancet
(1995)- et al.
Neural network modeling for surgical decisions on traumatic brain injury patients
International Journal of Medical Informatics
(2000) - et al.
Using neural networks to aid the diagnosis of breast implant rupture
Computers Operation Research
(1997) - et al.
Design of hybrid architectures based on neural classifier and RBF pre-processing for ECG analysis
International Journal of Approximate Reasoning
(1999) - et al.
Prediction of diagnosis in patients with early arthritis using combined Kohenen mapping and instance based evaluation criterion
Artificial Intelligence in Medicine
(2004) Fuzzy set theory in medical diagnosis
IEEE Transactions on Systems Man Cybernetic
(1986)Artifical neural networks to identify acute myocardial infarction—Reply
Lancet
(1996)- et al.
On the construction and training of reformulated Radial basis function neural networks
IEEE Transactions on Neural Networks
(2003)
Cited by (10)
Diagnosis of hypoglycemic episodes using a neural network based rule discovery system
2011, Expert Systems with ApplicationsCitation Excerpt :The advantages of using neural network approaches in diagnosis are their generalization ability in addressing both the nonlinear and fuzzy nature of the patients’ data. Although neural networks have been applied in building diagnosis models for various diagnoses like gastrointestinal disorders (Aruna, Puviarasan, & Palaniappan, 2007), abdominal pain (Mantzaris, Anastassopoulos, & Gardikis, 2008), urological dysfunctinos (Gil, Johnson, Chamizo, Paya, & Fernandez, 2009), dermatologic disease (Chang & Chen, 2009), breast cancer (Tanaka & Watada, 1998), heart disease (Das, Turkoglu, & Sengur, 2009), these diagnosis models only have the capability to transform the nonlinear or fuzzy patients’ data into simplified black-box structures in which no explicit knowledge or information can observed. Because of the black-box nature of the neural networks, some medical doctors may feel uncomfortable to use neural networks for diagnosis purposes even through the approaches may achieve better accuracy diagnosis than the other explicit modeling methods like classical statistical methods or genetic programming.
Bowel-sound pattern analysis using wavelets and neural networks with application to long-term, unsupervised, gastrointestinal motility monitoring
2008, Expert Systems with ApplicationsEffectiveness of Backpropagation Algorithm in Healthcare Data Classification
2021, Lecture Notes in Networks and SystemsAutomatic grading of adult depression using a backpropagation neural network classifier
2016, Data Mining in Biomedical Imaging, Signaling, and SystemsInterval type-2 fuzzy image processing expert system for diagnosing brain tumors
2014, 2014 IEEE Conference on Norbert Wiener in the 21st Century: Driving Technology's Future, 21CW 2014 - Incorporating the Proceedings of the 2014 North American Fuzzy Information Processing Society Conference, NAFIPS 2014, Conference Proceedings