Abstract:Screening for osteoporosis is crucial for early detection and prevention, yet it faces challenges due to the low accuracy of calcaneal quantitative ultrasound (QUS) and limited access to dual-energy X-ray absorptiometry (DXA) scans. Recent advances in AI offer a promising solution through opportunistic screening using existing medical images. This study aims to utilize deep learning techniques to develop a model that analyzes chest X-ray (CXR) images for osteoporosis screening. This study included the AI model development stage and the clinical validation stage. In the AI model development stage, the combined dataset of 5122 paired CXR images and DXA reports from the patients aged 20 to 98 years at a medical center was collected. The images were enhanced and filtered for hardware retention such as pedicle screws, bone cement, artificial intervertebral discs or severe deformity in target level of T12 and L1. The dataset was then separated into training, validating, and testing datasets for model training and performance validation. In the clinical validation stage, we collected 440 paired CXR images and DXA reports from both the TCVGH and Joy Clinic, including 304 pared data from TCVGH and 136 paired data from Joy Clinic. The pre-clinical test yielded an area under the curve (AUC) of 0.940, while the clinical validation showed an AUC of 0.946. Pearson's correlation coefficient was 0.88. The model demonstrated an overall accuracy, sensitivity, and specificity of 89.0%, 88.7%, and 89.4%, respectively. This study proposes an AI model for opportunistic osteoporosis screening through CXR, demonstrating good performance and suggesting its potential for broad adoption in preliminary screening among high-risk populations.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to predict lumbar bone mineral density (BMD) and T - scores from chest X - rays (CXR) by using deep - learning techniques, in order to achieve early screening for osteoporosis. Specifically, the paper aims to develop an artificial - intelligence - based software tool - VeriOsteoTMOP, which can utilize existing chest X - ray images to predict patients' lumbar bone mineral density and T - scores, thus providing a method that requires no additional radiation exposure, is low - cost and highly efficient for the preliminary screening of osteoporosis. ### Background problems 1. **Challenges in screening for osteoporosis**: - Osteoporosis is a chronic bone disease, characterized by low bone density and degradation of the bone tissue microstructure, leading to an increased risk of fractures. - Current bone density measurement methods, such as dual - energy X - ray absorptiometry (DXA), are accurate but the equipment is expensive and the penetration rate is low. - Another commonly used method, quantitative ultrasound of the calcaneus (QUS), is less expensive and portable, but has lower accuracy. 2. **Limitations of existing methods**: - The high cost and limited availability of DXA scans limit their application in large - scale screening. - Although QUS is cheap, its sensitivity and specificity are low and it is not sufficient as a primary screening tool. ### Objectives of the paper - **Develop an AI model**: Use deep - learning techniques to develop a model that can analyze chest X - ray images and predict lumbar bone mineral density and T - scores. - **Clinical verification**: Through the clinical verification stage, evaluate the performance of this AI model in the actual medical environment to ensure its accuracy and reliability. ### Method overview - **Data collection**: 5,122 pairs of chest X - ray images and DXA reports were collected from a medical center in central Taiwan for the development of the AI model. - **Model training**: DenseNet - 121 was used as the backbone network. The data was pre - processed through random affine transformation and image enhancement techniques, and the mean - squared - error (MSE) was used as the loss function for model training. - **Clinical verification**: 440 pairs of chest X - ray images and DXA reports were collected from two different medical institutions (TCVGH and Joy Clinic) for clinical verification. ### Main results - **Performance indicators**: The AUC value of the model in the clinical verification stage was 0.946, the Pearson correlation coefficient was 0.88, the overall accuracy rate was 89.0%, the sensitivity was 88.7%, and the specificity was 89.4%. - **Potential applications**: The research shows that this AI model exhibits good performance in the preliminary screening of osteoporosis in high - risk populations and has the potential for wide promotion. In conclusion, this paper provides a new, efficient and economical method for the early screening of osteoporosis by developing and verifying a deep - learning - based AI model.

Clinical Validation of a Deep Learning-Based Software for Lumbar Bone Mineral Density and T-Score Prediction from Chest X-ray Images