Final Report Summary - BONEMATCH (Bone Multi-modal Automated Trabecular Histomorphometry)
Bone Multi-modal Automated Trabecular Histomorphometry (BoneMATCH) aimed atinvestigating bone microarchitecture changes using high-resolution peripheral quantitative computed tomography (HRpQCT), and at discovering relationships between patterns of change and clinical factors such as aging, disease and organ transplantation.
The project was motivated by the long-term goal of developing a non-invasion “virtual bone biopsy” which enables clinical evaluation of bone microarchitecture through in-vivo imaging. BoneMATCH serves as a platform for developing clinical image analysis algorithms using state-of-the-art research imaging techniques such microCT and HRpQCT, and then migrating those techniques into clinical imaging such as MDCT through validation on patient cohorts. In this context, BoneMATCH made significant advances in serveral directions.
1. Data acquisition towards a multi-scale parallel corpus of bone microarchitecture
Data used for BoneMATCH was sourced using a combination of HRpQCT images from patient cohorts, healthy volunteers, and cadaver specimens. Additionally, we imaged cadaver specimens using micro CT (µCT), HR-pQCT as well as clinical multi-detector CT (MDCT). We implemented highly-accurate registration methods to align these data, and the resulting parallel corpus of imaging data covers bone microarchitecture from the scale of 4 microns to corresponding clinical imaging data. This corpus is the basis for multi-scale, multi-modal image analysis, and algorithm development and validation.
2. Algorithm development to capture changing bone microarchitecture
Following data acquisition, the focus of BoneMATCH was on algorithm development and integration for the purpose of developing and validating advanced clinical analysis methods in bone imaging. These methods included 3D texture-based trabecular bone quality assessment, localized longitudinal assessment of bone microarchitecture changes, dictionary learning for denoising and super resolution of HRpQCT images, and spatio-temporal mapping of clinical and microstructural parameters to localized bone microarchitectural changes.
Results from BoneMATCH include clinical validation of 3D texture-based trabecular bone quality maps (BQMs) in a cohort of lung transplant recipients. The BQM has been developed to provide radiologists and clinicians with information about bone microarchitecture which goes beyond bone mineral density (BMD), the gold standard for bone health, particularly in cases where images with similar BMD scores have a noticeably different structural appearance.
3. Super-resolution and denoising: briding scales and modalities
As a specific focus the BoneMATCH project has explored the use of compressed sensing and sparse signal reconstruction to apply denoising and super-resolution in HRpQCT imaging.
4. Clinically applicable algorithms and concise visualization of bone microarchitecture change
Finally, in order to enable the use of BoneMATCH methods in a research- and clinical context it has introduced the concept of the “bone morphogram”, a method for quantifying and visualizing bone microstructural changes which goes beyond the clinical gold standard of comparing BMD, cortical porosity and other summary values between time points.
BoneMATCH has produced a collection of software tools for continued research into bone microarchitectural assessment, as well as for establishing standards for advanced clinical quantitative imaging for bone health. The CIR lab at the Medical University of Vienna is actively working with internal and external collaborators to validate and expand on BoneMATCH algorithms on various patient cohorts, and to facilitate the continued use of the tools by the research community.
Through these follow-on studies to correlate our techniques with clinical and biomechanical parameters, the research conducted during BoneMATCH should lead to better fracture risk assessment, more detailed analysis of drug therapy response in osteoporosis and other bone diseases, as well as a reduction in radiation exposure to patients through image enhancement and advanced quantitative assessment.
The project was motivated by the long-term goal of developing a non-invasion “virtual bone biopsy” which enables clinical evaluation of bone microarchitecture through in-vivo imaging. BoneMATCH serves as a platform for developing clinical image analysis algorithms using state-of-the-art research imaging techniques such microCT and HRpQCT, and then migrating those techniques into clinical imaging such as MDCT through validation on patient cohorts. In this context, BoneMATCH made significant advances in serveral directions.
1. Data acquisition towards a multi-scale parallel corpus of bone microarchitecture
Data used for BoneMATCH was sourced using a combination of HRpQCT images from patient cohorts, healthy volunteers, and cadaver specimens. Additionally, we imaged cadaver specimens using micro CT (µCT), HR-pQCT as well as clinical multi-detector CT (MDCT). We implemented highly-accurate registration methods to align these data, and the resulting parallel corpus of imaging data covers bone microarchitecture from the scale of 4 microns to corresponding clinical imaging data. This corpus is the basis for multi-scale, multi-modal image analysis, and algorithm development and validation.
2. Algorithm development to capture changing bone microarchitecture
Following data acquisition, the focus of BoneMATCH was on algorithm development and integration for the purpose of developing and validating advanced clinical analysis methods in bone imaging. These methods included 3D texture-based trabecular bone quality assessment, localized longitudinal assessment of bone microarchitecture changes, dictionary learning for denoising and super resolution of HRpQCT images, and spatio-temporal mapping of clinical and microstructural parameters to localized bone microarchitectural changes.
Results from BoneMATCH include clinical validation of 3D texture-based trabecular bone quality maps (BQMs) in a cohort of lung transplant recipients. The BQM has been developed to provide radiologists and clinicians with information about bone microarchitecture which goes beyond bone mineral density (BMD), the gold standard for bone health, particularly in cases where images with similar BMD scores have a noticeably different structural appearance.
3. Super-resolution and denoising: briding scales and modalities
As a specific focus the BoneMATCH project has explored the use of compressed sensing and sparse signal reconstruction to apply denoising and super-resolution in HRpQCT imaging.
4. Clinically applicable algorithms and concise visualization of bone microarchitecture change
Finally, in order to enable the use of BoneMATCH methods in a research- and clinical context it has introduced the concept of the “bone morphogram”, a method for quantifying and visualizing bone microstructural changes which goes beyond the clinical gold standard of comparing BMD, cortical porosity and other summary values between time points.
BoneMATCH has produced a collection of software tools for continued research into bone microarchitectural assessment, as well as for establishing standards for advanced clinical quantitative imaging for bone health. The CIR lab at the Medical University of Vienna is actively working with internal and external collaborators to validate and expand on BoneMATCH algorithms on various patient cohorts, and to facilitate the continued use of the tools by the research community.
Through these follow-on studies to correlate our techniques with clinical and biomechanical parameters, the research conducted during BoneMATCH should lead to better fracture risk assessment, more detailed analysis of drug therapy response in osteoporosis and other bone diseases, as well as a reduction in radiation exposure to patients through image enhancement and advanced quantitative assessment.