[Paper Review] Expert System for Diagnosis of Chest Diseases Using Neural Networks

Ismail Kayali|arXiv (Cornell University)|Feb 19, 2018

Information Systems and Technology Applications4 references3 citations

TL;DR

This paper presents a neural network-powered expert system for diagnosing chest diseases, integrating forward and backward chaining with backpropagation-based learning to improve diagnostic accuracy despite overlapping symptoms. The system enhances clinical decision-making by supporting both novice practitioners and experienced doctors through rule-based inference, explanation generation, and treatment recommendations, with potential for use in hospitals and medical education.

ABSTRACT

This article represents one of the contemporary trends in the application of the latest methods of information and communication technology for medicine through an expert system helps the doctor to diagnose some chest diseases which is important because of the frequent spread of chest diseases nowadays in addition to the overlap symptoms of these diseases, which is difficult to right diagnose by doctors with several algorithms: Forward Chaining, Backward Chaining, Neural Network(Back Propagation). However, this system cannot replace the doctor function, but it can help the doctor to avoid wrong diagnosis and treatments. It can also be developed in such a way to help the novice doctors.

Motivation & Objective

• To develop an expert system that assists doctors in diagnosing chest diseases with high symptom overlap.
• To integrate neural networks with rule-based inference (forward and backward chaining) to improve diagnostic accuracy.
• To support novice medical practitioners and serve as a teaching tool in medical education.
• To enable the system to suggest treatments based on patient-specific constraints and disease diagnosis.
• To lay the foundation for future expansion to include a broader range of chest diseases through data mining and larger datasets.

Proposed method

• The system employs a hybrid inference mechanism combining forward and backward chaining to balance user input requirements and diagnostic completeness.
• Neural networks (backpropagation) are used to classify and separate overlapping symptom patterns in chest disease diagnosis.
• Production rules represent medical expertise, with each rule structured as IF-THEN conditions and actions using variable-operator-value triples.
• The system includes six core components: knowledge base, inference engine, explanation subsystem, knowledge-base editor, administration interface, and management system.
• A database stores all permanent and temporary data, supporting networked access for collaboration across clinics and institutions.
• The system is designed with web deployment in mind, enabling access via local networks or the internet for scalable use.

Experimental results

Research questions

• RQ1 Can a hybrid expert system combining rule-based inference and neural networks improve the accuracy of chest disease diagnosis despite overlapping symptoms?
• RQ2 How effectively can neural networks classify complex, convergent symptom patterns in chest diseases compared to traditional rule-based systems?
• RQ3 To what extent can such a system support novice doctors and serve as a teaching tool in medical education?
• RQ4 What is the impact of integrating data mining and machine learning on the system's ability to discover new diagnostic patterns?
• RQ5 How can the system be extended to cover a broader range of chest diseases with improved generalization using larger datasets?

Key findings

• The expert system successfully integrates neural networks with rule-based inference to improve diagnostic accuracy for chest diseases with overlapping symptoms.
• The hybrid chaining approach allows flexible user interaction by prompting for data as needed, rather than requiring all data upfront.
• The explanation subsystem enhances transparency by justifying diagnostic conclusions, increasing user trust and system usability.
• The system demonstrates strong potential as a clinical decision support tool in hospitals and medical centers.
• The system can be used to archive and analyze clinical cases, enabling future data mining for discovering new diagnostic patterns.
• The authors identify the need for larger, more diverse datasets to fully realize the system's potential in covering all chest diseases.