Radiomics-based classification of medication-related osteonecrosis of the jaw using panoramic radiographs.

Abstract

Objectives: Medication-related osteonecrosis of the jaw (MRONJ) caused by bone resorption inhibitors is difficult to treat and reduces the patient's quality of life. We aimed to classify the likelihood of MRONJ development using panoramic radiographs taken prior to bone resorption inhibitor administration.

Methods: We included patients who underwent panoramic radiographic evaluation at Hiroshima University Hospital prior to bone resorption inhibitor administration. Thirty-two patients with MRON of the mandible (16 men and 16 women) and 57 without MRONJ (23 men and 34 women) were selected. The mandible was segmented from the mental foramen to the anterior mandibular angle notch on panoramic radiographs before treatment. The image features within this region were extracted and quantified. Overall, 13 shape, 18 histogram-based, 75 texture-based, and 744 wavelet features were extracted. Least absolute shrinkage and selection operator regression were used to select relevant features from the extracted data. Support vector machine (SVM) and neural network of multilayer perceptron (MLP) were used as machine-learning models. The sensitivity, specificity, and area under the curve (AUC) from the receiver operating characteristic curves were used to evaluate diagnostic performances.

Results: The SVM model achieved a sensitivity of 0.667, a specificity of 0.833, and an AUC of 0.903 in the test group. Meanwhile, the MLP model achieved a sensitivity of 0.833, a specificity of 0.750, and an AUC of 0.903 in the test group.

Conclusion: Radiomics analysis of panoramic radiographs taken before bone resorption inhibitor administration can differentiate between patients with MRONJ and those without MRONJ.