The early detections of osteoporosis and osteopenia are required to avoid the painful and life-altering bone fractures, but the screening rate is still limited all over the world. Therefore, to detect and alert people to their lower bone mineral densities (BMDs), the accessible and easy methods are needed. In this study, we developed the fast-screening method for osteoporosis by using chest X-ray images taken frequently and then evaluated the performance of proposed method. We used both BMD values measured by dual-energy X-ray absorptiometry (DXA) and chest X-ray images from 711 females. In the proposed method, by using deep convolutional neural network (DCNN), images were classified into normal BMD cases and lower BMD cases. DCNN was trained by ROI images which are cropped first lumber spine from chest X-ray images. The sensitivity, specificity, overall accuracies and AUC were respectively 87.95%, 79.60%, 84.18% and 0.9134. We developed and validated the osteoporosis screening algorithm based on DCNN by using chest X-ray images. The proposed system has high potential as a classification tool, and there is a possibility that the vertebral bodies on chest X-ray images show characteristic of lower BMD.