Osteoporosis Prescreening and Bone Mineral Density Prediction using Dental Panoramic Radiographs

Area of Science:

• Radiology
• Artificial Intelligence
• Osteoporosis Research

Background:

• Dental Panoramic Radiographs (DPR) show potential for osteoporosis screening due to bone density and structure correlations.
• Limited datasets pose challenges for traditional deep learning models in osteoporosis detection.
• Existing research often relies on pretrained models for feature extraction and classification.

Purpose of the Study:

• To evaluate a deep convolutional neural network (DCNN)-based computer-assisted diagnosis (CAD) system for osteoporosis detection in DPR images.
• To compare the CAD system's diagnostic performance against oral and maxillofacial radiologists.
• To develop a regression model for predicting Bone Mineral Density (BMD) using DPR images, patient age, and osteoporosis labels.

Main Methods:

• A multi-column DCNN model was developed for osteoporosis detection using a dataset of 70 DPR images.
• Computer vision techniques and patient age metadata were employed to enhance model performance.
• A neural network-based regression model was created to predict BMD values.

Main Results:

• The DCNN-based CAD system achieved 87.86% accuracy in detecting osteoporosis.
• The system's performance demonstrated high accord with diagnoses made by experienced radiologists.
• The study successfully leveraged DPR images, age data, and labels to predict BMD.

Conclusions:

• DPR images, combined with patient age data and a DCNN CAD system, offer a viable method for osteoporosis detection.
• The developed system shows comparable diagnostic performance to expert radiologists.
• This approach facilitates osteoporosis prescreening and BMD value prediction, aiding in patient management.