Machine learning to predict the occurrence of bisphosphonate-related osteonecrosis of the jaw associated with dental extraction: A preliminary report

Shan

et al. 2020

Preprint

Background: To investigate the effectiveness and feasibility of machine learning models based on trabecular microstructure parameters for predicting the occurrence of marginal bone loss (MBL) of the submerged dental implant in mandible. Methods: Clinical variables and morphological parameters of trabecular bone were collected from 81 subjects with submerged implants in the mandible (41 cases of abnormal MBL and 40 as normal controls). We measured the peri-implant MBL level by a cone-beam computed tomography (CBCT) at the follow-up of 20.95±2.67 months after functional loading. The morphological parameters and possible factors associated with MBL were collected in a mean of 3.98±1.06 months at the early loading stage. All variables were analyzed using correlation and covariance matrices. Support vector machine (SVM), artificial neural network (ANN), logistic regression (LR) model and random forest (RF) were actualized to predict abnormal MBL.Results: At the early stage of functional loading, the abnormal MBL cases showed a significant increase of structure model index (SMI) and trabecular pattern factor (Tb.Pf) in peri-implant. Meanwhile, SMI and Tb.Pf simultaneously revealed a significantly high positive correlation with MBL. The SVM model exhibited the best outcome in predicting MBL (AUC = 0.967), followed by ANN (AUC = 0.928), LR (AUC = 0.906), RF (AUC = 0.842), SMI alone (AUC = 0.705) and Tb.Pf alone (AUC = 0.663). Compared with one single predictor, all algorithm models yielded significantly superior performance.Conclusion: Abnormal MBL cases demonstrated the premonitory morphological variation in trabecular bone at the early stage. MBL prediction could be achieved by machine learning methods.

Section: Discussionmentioning

confidence: 99%

“…Machine learning has been gradually adopted to predict the progression of disease. [15,16] To the best of our knowledge, using machine learning to predict MBL has rarely appeared in published studies.…”

Section: Introductionmentioning

confidence: 99%

Trabeculae microstructure parameters serve as effective predictors for marginal bone loss of dental implant in the mandible 

Shan

et al. 2020

Preprint

“…Machine-learning algorithm has been presented recently to predict the occurrence of MRONJ following dental extraction in patients on bisphosphonate therapy [29]. The performance of machine-learning models was compared along with single predictors such as serum CTX level.…”

Section: Discussionmentioning

confidence: 99%

Value of pre-operative CTX serum levels in the prediction of medication-related osteonecrosis of the jaw (MRONJ): a retrospective clinical study

et al. 2019

Objectives The low incidence yet severe presentation of medication-related osteonecrosis of the jaw (MRONJ) makes it necessary to develop reliable predictive and preventive strategies. This study explored the value of pre-operative carboxy-terminal collagen crosslinks (CTX) serum level in the prediction of osteonecrosis-related complications in patients on bisphosphonate therapy. Patients and methods We examined patient records over 4 years (a total of 137 patients). Biometric data were extracted, in addition to type of treatment, CTX levels, drug holiday, procedure, complications, and co-morbidities. Non-parametric Wilcoxon two-sample tests were used to test the effect of initial CTX level in IVor PO and whether it was predictive of complications. Two independent proportion tests were used for testing the two different complication incident rates before or after the drug holiday. Results A total of 93 patients were included in the study, of whom 88.17% were female. A total of 11 patients were receiving IV bisphosphonates at the time of initial presentation, 82 oral bisphosphonates. Out of 64 patients who underwent invasive dental procedure (IDP) before a drug holiday, eight were on IV bisphosphonates. Three patients in this group experienced osteonecrosisrelated complications (37.5%). Out of the remaining 56 patients on oral bisphosphonates, four (7.14%) developed complications, significantly lower than the IV bisphosphonate group (p = 0.0364). On the other hand, of the 34 patients placed on a drug holiday prior to IDP, only one subject developed complications related to osteonecrosis. Five subjects who had operations both before and after drug holiday did not experience any complications. No statistical difference was detected in complication rates based on initial CTX level (above versus below 150 pg/ml), gender, comorbidities, or total duration of bisphosphonate treatment (p = 0.2675). The sensitivity and specificity of CTX cutoff of 150 pg/ml in predicting osteonecrosis were 37.5% and 57.7, respectively. Conclusions Serum levels of CTX by itself are not reliable as a predictive or preventive measure for such complications. Our data also suggested that a drug holiday of 5 months was not helpful in preventing osteonecrosis-related complications in patients on intravenous bisphosphonates. Further studies are urgently needed to develop adequate predictive and preventive strategies of MRONJ.

Proteomics Clinical Apps

“…These errors are back‐propagated through the neural network, and the weights of the connections between the neurons are updated to minimize this error. After the training phase, including multiple iterations of forward and back propagation, the performance of the CNN is assessed by an unseen test data set . This process is illustrated in Figure .…”

Section: Concepts Of Ai ML and Dlmentioning

confidence: 99%

Radiomics and Machine Learning in Oral Healthcare

Leite

Vasconcelos

Willems

et al. 2020

The increasing storage of information, data, and forms of knowledge has led to the development of new technologies that can help to accomplish complex tasks in different areas, such as in dentistry. In this context, the role of computational methods, such as radiomics and Artificial Intelligence (AI) applications, has been progressing remarkably for dentomaxillofacial radiology (DMFR). These tools bring new perspectives for diagnosis, classification, and prediction of oral diseases, treatment planning, and for the evaluation and prediction of outcomes, minimizing the possibilities of human errors. A comprehensive review of the state-of-the-art of using radiomics and machine learning (ML) for imaging in oral healthcare is presented in this paper. Although the number of published studies is still relatively low, the preliminary results are very promising and in a near future, an augmented dentomaxillofacial radiology (ADMFR) will combine the use of radiomics-based and AI-based analyses with the radiologist's evaluation. In addition to the opportunities and possibilities, some challenges and limitations have also been discussed for further investigations.