Image Processing Pipeline to Improve the Detection of Vertical Root Fractures in Digital Periapical Radiographs

Abstract

This paper proposes an image processing pipeline to improve the detection of vertical root fractures (VRF) in digital periapical radiographs. The pipeline is composed by three steps: geometric adjustment, Gamma correction and adaptive contrast enhancement. The proposed method was compared to the conventional pre-processing pipeline normally applied to digital radiographs. Four experienced dentists (two radiologists and two endodontists) classified a set of digital periapical images according to the presence or absence of fracture, when images were pre-processed using both methods. Inter-examiner reproducibility was verified by the weighted Kappa test. Sensitivity, specificity and accuracy were calculated along the area under the receiver operating characteristic (ROC) curve for both methods. Results indicated that the proposed method increased the specificity of radiologists by 13% and the endodontists by 4% when the proposed pipeline was used. The proposed method yielded better results on the detection of VRF in digital periapical radiography images when compared to the traditional method.