Employing Cluster- and Threshold-Based Techniques for Lesion Interaction in Dental Radiography

Abstract

An image-processing-based technique for identifying and segmenting lesion regions in dental X-ray images is described in this research. The low contrast of these images and the similarity in intensity between lesion sites and surrounding tissues continue to make accurate detection difficult. Image enhancement approaches first improve image quality and highlight the lesion spots to address this issue. To enhance the visibility of important features, this work employs preprocessing techniques such as contrast adjustment and histogram equalization. It differentiates the similar homogeneous areas in the image that use different segmentation techniques like K-means clustering, fuzzy C-means (FCM), etc. The paper aims at suggesting a hybridization of fuzzy C-means and histogram equalization to improve the accuracy of the segmentation process. The results showed that with all hybrid methods, the computation time was decreased compared to the FCM algorithm, while the accuracy for lesion detection increased. Morphological operations are furthermore used to improve the quality of the obtained segmentation and retrieve the area of interest. The experimental results on dental X-ray images demonstrate that the proposed method can successfully localize lesions and boost segmentation performance.