Speckle reduction and lesion segmentation of OCT tooth images for early caries detection.

The significance of identifying early non-cavitated carious lesions and monitoring the lesion extent has led to increasing prospects for prevention, early diagnosis, and implementation of conservative treatments. This paper emphasizes the importance of speckle reduction and possible lesion segmentation options of optical coherence tomography (OCT) images prior to caries detection. First, a comparison of popular speckle reduction filters is presented. These filtering algorithms were evaluated to measure the ability of different methods for reducing background noise from raw images. Both qualitative and quantitative results (signal-to-noise ratio, contrast-to-noise ratio) are reported. Image segmentation is then applied to multiple tooth images. With proper thresholding, high intensity response regions are outlined with the possibility of assessing caries and monitoring its regression. Our results show that a rotating kernel transformation (RKT) filter with 9x9 kernel size provides a good compromise between noise reduction yet preserving the pathological features of interest as required for subsequent feature segmentation analyses.