Volume 11, Issue 4 (3-2025)
jhbmi 2025, 11(4): 306-322 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Maghsoudi R, Tavan M. Differential Diagnosis of Precancerous and Cancerous Oral Lesions Based on an Integrated Artificial Intelligence Model. jhbmi 2025; 11 (4) :306-322
URL: http://jhbmi.ir/article-1-902-en.html
URL: http://jhbmi.ir/article-1-902-en.html
Department of Computer Engineering, Nur Branch, Islamic Azad University, Nur, Iran
Abstract: (198 Views)
Introduction: Early detection of precancerous and cancerous oral lesions is essential for improving treatment outcomes and reducing patient mortality. However, the similarity of these lesions to benign lesions complicates the diagnostic process. Computer-aided diagnostic (CAD) systems can enhance diagnostic accuracy and reduce the time required for diagnosis. This study aimed to develop an integrated model for the diagnosis of oral lesions
Method: In this study, clinical images of patients with leukoplakia, lichen planus, and oral squamous cell carcinoma were collected. After preprocessing the images to improve quality and remove noise, key features were extracted using the SURF algorithm. To reduce data dimensions and select effective features, the K-means clustering algorithm was employed, resulting in a reduction to 30 features. Image classification was performed using five machine learning algorithms: support vector machine (SVM), multilayer neural network (MLP), radial basis function (RBF), decision tree, and Bayesian classifier. The performance of these algorithms was evaluated comparatively.
Results: The results indicated that SVM excelled in detecting oral lesions, achieving an accuracy of 95%. This algorithm demonstrated significant superiority over other methods due to its ability to manage high-dimensional data and its capabilities for both linear and nonlinear separation. While MLP and RBF also provided acceptable results, their accuracy was lower than that of SVM. Employing K-means for dimensionality reduction improved both the speed and accuracy of classification.
Conclusion: This study presents, for the first time, an integrated model for high-accuracy diagnosis of oral lesions using imaging. This approach minimizes misdiagnosis errors and reduces the time and costs associated with diagnosis. The application of artificial intelligence in diagnosing oral lesions can enhance healthcare quality as an auxiliary tool alongside dental professionals. This model holds potential for development in other medical fields and can serve as a reference for future research.
Method: In this study, clinical images of patients with leukoplakia, lichen planus, and oral squamous cell carcinoma were collected. After preprocessing the images to improve quality and remove noise, key features were extracted using the SURF algorithm. To reduce data dimensions and select effective features, the K-means clustering algorithm was employed, resulting in a reduction to 30 features. Image classification was performed using five machine learning algorithms: support vector machine (SVM), multilayer neural network (MLP), radial basis function (RBF), decision tree, and Bayesian classifier. The performance of these algorithms was evaluated comparatively.
Results: The results indicated that SVM excelled in detecting oral lesions, achieving an accuracy of 95%. This algorithm demonstrated significant superiority over other methods due to its ability to manage high-dimensional data and its capabilities for both linear and nonlinear separation. While MLP and RBF also provided acceptable results, their accuracy was lower than that of SVM. Employing K-means for dimensionality reduction improved both the speed and accuracy of classification.
Conclusion: This study presents, for the first time, an integrated model for high-accuracy diagnosis of oral lesions using imaging. This approach minimizes misdiagnosis errors and reduces the time and costs associated with diagnosis. The application of artificial intelligence in diagnosing oral lesions can enhance healthcare quality as an auxiliary tool alongside dental professionals. This model holds potential for development in other medical fields and can serve as a reference for future research.
Keywords: Artificial intelligence, Machine learning algorithms, Classification, Feature extraction, Oral lesions.
Type of Study: Original Article |
Subject:
Artificial Intelligence in Healthcare
Received: 2024/11/25 | Accepted: 2025/03/1
Received: 2024/11/25 | Accepted: 2025/03/1
Audio File [MP3 926 KB] (5 Download)
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
