Vertebral fractures are the most common osteoporotic fractures, with a prevalence of 12-20% in Europe, making them a major health problem because of the associated morbidity, mortality and costs. Without adequate treatment, vertebral fractures are often followed by subsequent fractures, leading to further invalidation and deterioration of health. Therefore, early detection of vertebral fractures is important so preventative treatment can be initiated.

Vertebral fracture assessment (VFA) using Dual-Energy X-ray Absorptiometry (DXA) equipment is an imaging technique in which a lateral image of the spine is made. With vertebral morphometry, vertebral heights are measured and fractures are identified when vertebral height is lower than expected. However, vertebral morphometry is time and labor-intensive, and is subject to inter-operator variability. Automating VFA using Artificial Intelligence (AI) could help to overcome these limitations. We employed a co-development approach, aiming to create an AI-based tool to automatically perform vertebral morphometry on VFA images to identify vertebral fractures.

Firstly, we conducted a literature review to investigate the current use of AI in quantitative DXA imaging in a broad sense (Chapter 2). Besides VFA, other quantitative parameters describing bone macrogeometry and microgeometry can be extracted from DXA images. Incorporating these risk factors into multivariate prediction models could improve the identification of those at risk of fracture and help in clinical decision making. Although still in development, AI has been successfully applied in aid of fracture risk assessment, showing promising results.

In Chapter 3, the results of our reader study are described, evaluating VFA with manual vertebral morphometry as it is currently performed. This study served as a baseline measurement to quantify the effort needed to perform manual VFA and assess the potential value of automating VFA. The average annotation time per VFA image was 259 seconds. Although the intraclass correlation for vertebral height measurements between different readers was high, inter-observer agreement for fracture classification was only poor to moderate.

Together with our industry partner, we developed an AI-based software tool to perform vertebral morphometry. This tool is still in development, and we evaluated its current performance and potential impact in the study described in Chapter 4. Although its current standalone performance is suboptimal and shows room for improvement, this initial investigation showed that automated VFA has the potential to significantly reduce the required reader time.

Finally, in Chapter 5 we reflect on our user-centered co-development process and the next steps to bring our AI tool to market. We believe that collaboration between academic healthcare institutions and industry are essential for successful development of AI products. Validation of these products throughout the development process and actively involving intended users should be done. In the near future, vertebral fracture assessment can be supported by our AI-based application, potentially leading to lower annotation times and improving clinical workflow. However, further improvements have to be made and independent validation is required before market access.

Introduction: In daily practice, at interventional radiology (IR), percutaneous needle placement (e.g. biopsy and ablation) is generally performed under real time guidance of morphological imaging (US or CT). However, in some cases obtaining representative specimen can be challenging causing repeated biopsies or shift to open procedures. And so alternatively, needle placement can occur using computer-assisted needle navigation strategies that rely on pre-interventional images such as (PET-) CT scans. To unravel the impact of needle navigation strategies in addition to the use of ultrasound guidance, we recorded and analyzed the needle- and US-movements during exercises performed in a custom abdominal biopsy phantom.

Methods: custom abdominal biopsy phantom was generated using synthetic ballistic gel, bone-like structures, spherical targets and fiducials for both electromagnetic (EM) and optical tracking. Following administration of a mixture containing 18F-FDG and 99mTcO4- and prior to the needle interventions, the phantom was subjected to PET-CT and SPECT-CT imaging. After presenting them with the 3D imaging data-sets interventions were performed by both novices and experts (n = 20 in total). The exercise comprised of three consecutive steps: 1) US guided biopsy, 2) US guided PET-CT-navigated biopsy (US + Nav), 3) US guided PET-CT-navigated biopsy + virtual needle guidance (US + Nav + NG). Each step was performed on a different target and the exercise continued until the participant was successful. In all cases, we recorded the traveled paths determined in 3D (i.e., x, y, and z) of both the needle and the ultrasound probe. Tracking was realized by using a fiducial-based optical tracking system. Following the visualization of the traveled paths in MATLAB® (the MathWorks, Inc), we used custom software to analyze movement features such as speed, acceleration, jerkiness, straightness index, angular dispersion and curvature and related these dexterity (Dx) scores. We also analyzed missteps such as corrections and retractions and combined these with dexterity features to create decision making (DM) scores.

Results: Analysis of the needle trajectories indicated a decrease of jerkiness for the novices as a result of the additional navigational technologies (US vs Us + Nav + NG; p = 0.033). In this particular group US + Nav influenced both the dexterity and decision making more positively than US + Nav + NG (Dx80(US) = 20.0 (US) vs Dx80(US+Nav) = 5.19 vs Dx80(US+Nav+NG) = 9.73 and Dm80(US) = 3.03 vs Dm80(US+Nav) = 0.446 vs Dm80(US+Nav+NG) = 1.24). For experts, the navigational technologies negatively influenced the dexterity of experts similar but reversed with scores of Dx80(US) = 2.92 vs Dx80(US+Nav) = 12.1 vs Dx80(US+Nav+NG) = 9.33 and negatively influenced their decision making by increasing total pathlength, corrections and retractions (US vs US + Nav; p = 0.035, p = 0.001 and p = 0.006, respectively), thus worsening decision making scores (Dm80(US) = 0.0259 vs Dm80(US+Nav) =1.06 vs Dm80(US+Nav+NG) = 1.26).

Conclusion: By recording and analyzing the movement paths of the ultrasound probe and biopsy needle trajectories it has become possible to make an objective value-assessment on the performance enhancement created by computers-assisted navigation strategies. These preliminary findings suggest that technological refinements have a different impact on novices and experts and that increasing the technological complexity can reduce its impact.