A Novel Morphological Analysis of DXA-DICOM Images by Artificial Neural Networks for Estimating Bone Mineral Density in Health and Disease

Document Type



One of the best methods for diagnosing bone disease in humans is site-specific and total bone mineral density (BMD) measurements by Dual-energy X-ray Absorptiometry (DXA) machines. The basic disadvantage of this technology is inconsistent BMD measurements among different DXA machines from different manufacturers due to different image analysis algorithms. The objective of the present study was to apply artificial neural networks (ANNs) to estimate total BMD for diagnosing a population of Egyptians with and without pathology, using extracted features from DXA-DICOM images based on the Histogram and Binary algorithms as compared to reference BMD measurements by DXA machine. The sample size comprised 3000 male and female participants with an age range 22–49 years, who were referred to us for diagnosis and/or treatment and for DXA total body scans in the period from January 2016 till December 2017. We constructed an entry computer data-logging visible unit, where we applied morphological operations to get a specific bone image, and used their extracted feature vectors as inputs to ANNs with cascade training, gathering, and testing for DXA-DICOM image processing. The multilayer feed-forward ANN set up its initial weights, carried out training and initiated the recall mode, and finally observed its decision and interaction based on estimated BMD. The ANN construction was carried out using a 3-layer architecture, with one hidden layer of 85 neurons. The input layer has neuron numbers equal to 256 for the Histogram and 77,365 for Binary algorithms, respectively. Total BMD estimation performance based on the Binary algorithm was capable of identifying all DXA-DICOM images with an accuracy of 100% for the training, cross-validation, and testing of the ANN phases. We believe this strategy will represent the means for standardizing bone measurements of all DXA machines, regardless of the manufacturer.

Publication Date



Faculty of Applied Health Sciences Technology

Subject Area

Health Sciences, General Medicine

Indexed in Scopus


Indexed in Web Of Science







Artificial neural network (ANN), bone mineral density (BMD), cascade training, digital imaging and communications in medicine (DICOM), dual X-ray absorptiometry (DXA), gathering and testing (CTGT), osteoporosis