This study proposes the use of a large population-based dataset of plain X-ray’s and confirmed Dual-energy X-ray absorptiometry (DXA) measured Bone mineral density (BMD) values to train a new deep learning model for diagnosing osteoporosis. It addresses the practical potential of being able to estimate BMD from X-ray’s in comparison to the financial limitations of relying on a DXA measurement to calculate BMD for an osteoporosis diagnosis. This research was done by collecting DXA and X-Rays of 4159 participants in Vietnam, then areal BMD (aBMD) data was collected from the femoral neck and the lumbar spine using DXA. Plain X-rays were then collected and the data from DXA was used with the X-ray data to train multiple AI models to detect a region of interest and then calculate the BMD (xBMD). The AI models were found to be mostly accurate, with r = 0.90 for femoral neck and r = 0.87 for lumbar spine BMD comparisons. It was also able to accurately predict fracture risks and has potential clinically for diagnosing both osteoporosis and fracture risk patients. While DXA remains the gold standard, AI analyzed X-ray data has the potential to be a valuable tool for low to middle income countries with resource limitations for CT or MRI as well as of course DXA.