| Summary: | This study presents a Convolutional Neural Network (CNN) model developed to address the limitations of current diagnostic methods for oral ulcers. Specifically, it targets the classification of three types of oral ulcers: Oral Malignant Ulcer, Recurrent Aphthous Ulcer, and Traumatic Ulcer. Traditional diagnostic approaches often suffer from limitations such as subjective interpretation, variability in diagnostic accuracy, and inadequate automation. To overcome these challenges, our approach involves constructing a CNN from scratch and rigorously optimizing key hyperparameters, including batch size and learning rate. Under optimal conditions, with a batch size of 12 and a learning rate of 0.001, our CNN achieved an accuracy of 98.61% and a loss rate of 4.78%. In contrast, alternative configurations resulted in an accuracy of 90.62% and a loss of 17.12%, highlighting the model's sensitivity to parameter adjustments. These results demonstrate that careful tuning of CNN hyperparameters can significantly improve diagnostic performance and accuracy. The findings suggest that the proposed CNN model offers a promising enhancement over traditional methods, with the potential to improve diagnostic precision and clinical outcomes for oral ulcers. Future work will focus on validating the model in real-world clinical settings and exploring its applicability to other medical imaging domains, with the aim of broadening its impact and utility in healthcare diagnostics. © 2024 IEEE.
|
|---|