A novel convolutional neural network–Fuzzy-based diagnosis in the classification of dental pulpitis

Abstract

Introduction: This study presents a computer-aided decision-making system based on the convolutional neural network (CNN)–fuzzy approach. According to the literature, there is a lack of coherence amongst dentists in diagnosing reversible or irreversible pulpitis. As a result, the goal of this research is to assist dentists in accurately diagnosing pulpitis. Materials and Methods: A rigorous algorithm that relies on CNN-fuzzy logic has been designed to handle inaccurate and ambiguous values of dental radiographs, as well as signs and symptoms of pulpitis. To begin, the probability of cavity for each class was determined using an independently designed CNN approach, which was then applied in combination with symptoms associated with pulpitis to a fuzzy knowledge base with 665 rules and the Mamdani inference algorithm to diagnose pulpitis and make recommendations to the dentist. Results: The CNN-fuzzy approach's results are compared to the dentists' recommendations. With the assistance of five professional dentists, the sensitivity, specificity, precision, accuracy, f1 score and Matthews correlation coefficient are calculated from 100 randomly generated sample cases. The CNN-fuzzy approach has a 94% accuracy, which is 7% higher than expert prediction. It is observed that the proposed approach produces results that are consistent with the dentists' diagnoses. Conclusion: The accuracy of the proposed computer-aided decision-making system for pulpitis increases dentists' confidence in diagnosing reversible and irreversible pulpitis and reduces false diagnoses due to ambiguous values of dental radiographs, signs and symptoms.