Probing long bones using ultrasound with machine learning: signal processing, simulation, inversion, and pattern recognition

点击翻译按钮获取中文摘要

Osteoporosis is one of the most common skeletal disorders characterized by reduced bone mass, micro-structural deterioration, and cortical bone thinning, leading to bone fragility and fracture risks. The disease can occur in either gender at any age, but women after menopause are more susceptible to the disease. The bone loss occurs silently without any symptoms until the bone is weak to break. Osteoporosis causes more than 8.9 million fractures annually worldwide and remains one of the most prevalent undiagnosed disease. Most patients, who experience osteoporotic fractures, lose their quality of life and especially the will to move freely. Therefore, preventing, being able to diagnose accurately, and monitoring osteoporosis and fragility fractures are of supreme importance. Currently, bone integrity is mainly assessed by measuring bone mineral density (BMD) using dual-energy X-ray absorptiometry (DXA). However, clinical studies indicated that BMD values measured in normal and osteoporotic bones showed considerable overlap; therefore, BMD does not provide sufficient information for accurate diagnosis of osteoporosis because X-ray based measurements fail to consider bone structural and mechanical properties. Among all methods, quantitative guided wave ultrasound (QGWU) has demonstrated its promising potential in the ex-vivo and in-vivo assessment of bone integrity. An advantage of QGWU over the gold-standard DXA is its ability to provide measurements of mechanical properties, relevant to bone strength. The cortex, which is sandwiched between soft tissue and marrow, acts as an ultrasound waveguide. The ultrasound trapped within the cortex reverberates and reinforces into a "bundle" of energetic ultrasonic guided waves (UGW) energy to travel along the long bone. UGW are sensitive to material thickness and elastic properties and thus, have the potential to probe the geometric and material properties of the cortex of long bones. The proposal offers an opportunity to develop innovative and smart ultrasound-based imaging technologies to discriminate healthy and osteoporotic populations and to diagnose and monitor the progression of bone diseases, which will improve patient care. The work here provides an excellent environment to train highly qualified personnel. There is a strong emphasis on cross-fertilization of several disciplines (medical physics, biomedical engineering, geophysics, computer science) to solve a medical-relevant problem. The findings of this research plan will lead to new and smart methodologies and technology development for ultrasound characterization of skeleton competence and integrity. Software development and design of transducer array to better measure cortical thinning and bone stiffness relevant to bone strength will lead to possible opportunities for patent and technology commercialization using ultrasonic guided waves.