Osteoporosis causes a reduction in bone density, which can result in bone fractures, particularly in the spine, forearms, and the hip. Osteoporosis does not present any symptoms, and yet it is estimated that more than 20 million people have this debilitating disease. Therefore, regular examination, especially in patients over the age of 65, is recommended.

Diagnosis is typically done by bone density measurement using Dual energy x-ray absorptiometry (DEXA) examination. DEXA is currently considered to be the gold standard for the diagnosis of osteoporosis. Even with 50 million scans per year, a large portion of metastases and fractures remain unnoticed by radiologists. This can lead to increased treatment cost and morbidity downstream.

Advances in computing have created numerous opportunities for artificial intelligence (AI) to solve clinically important imaging problems, with promising applications that would benefit patients. Most notably, AI outperformed radiologists in diagnosis of pneumonia from chest X-rays. 

The inventors developed an AI algorithm using random forest classifiers to assist radiologists in measuring bone mineral density. The algorithm can be implemented on current scanners. Experimental data shows improvement in sensitivity (77.8%), specificity (100%), and accuracy (98%) of lumbar bony metastasis detection, and improvement in sensitivity (61%), specificity (91%), and accuracy (85%) of lumbar fracture detection, which can reduce treatment cost and offer significant benefits to patients.

• Can be easily and inexpensively installed and implemented on existing machines.
• High accuracy (98%), sensitivity (77.8%) and specificity (100%) in detecting bony metastases.
• High accuracy (85%) sensitivity (61%) and specificity (91%) in detecting factures.
• Can reduce cost of care by better identifying bony metastatic disease and/or fractures.

• Bone mineral density measurement
• Bony metastasis and fracture detection from dual x-ray absorptiometry (DEXA) exams