ULtrasonic Assessment of Bone Using Neural Networks

Osteoporosis afflicts over 20 million people in the US, responsible for more than 275,000 hipfractures annually. Currently, the primary means for assessment relies on densitometric techniques.These methods subject the patient to ionizing radiation, are relatively expensive, and do not alwaysprovide good estimates of bone strength. Ultrasound offers several potential advantages. It isnon-ionizing and relatively inexpensive. Moreover, since ultrasound is a mechanical wave and interactswith bone in a fundamentally different manner than electromagnetic radiation, it may be able to providemore accurate estimates of bone strength compared with current densitometric methods. The goal ofthis research is to significantly improve the effectiveness of current ultrasonic bone assessmenttechniques by demonstrating the feasibility of a powerful new technology, i.e. neural networks, whichcan be used in conjunction with ultrasonic measurements, to accurately determine bone density andstrength. The specific aims are to measure ultrasonic velocity and attenuation in bovine trabecular boneand use neural networks to determine its bone mineral density and ultimate strength.