Decision Tree Algorithm for the Classification of Dental Caries Severity via Saliva

Abstract

Dental caries is one of the most prevalent chronic diseases. Early detection is prominent to avoid the tooth weakening or worst the tooth loss. UV absorption spectroscopy is a non-invasive technique used for the detection of salivary alpha-amylase which are increasing in the presence of caries. Spectrum acquired from patient at Faculty of dentistry, UKM showed significant peak around 260-300 nm which are correspond to the absorption of amino acid found in salivary alpha-amylase. The spectra are preprocesses using autoscale and multiplicative scatter correction (MSC) to optimize the signal. Decision tree algorithm was implemented on the UV absorption spectra. The best model of decision tree obtained when using autoscale preprocessing method. The accuracy, precision, sensitivity and specificity for the validation data obtained were 0.65, 1.00, 0.29 and 1.00 respectively. The decision tree requires more tuning for the robustness for future application.