Ultrasound shear-wave visco-hyperelastography for prostate cancer localization

Prostate cancer is the most commonly diagnosed type of cancer among men in nearly two-thirds of the world. According to the latest global statistics, there were 1.47 million new cases and 397,000 deaths in 2022, with projections rising to 2.63 million cases and 823,000 deaths by 2045. Given its high prevalence and significant health impacts, prostate cancer diagnosis should be accurate, noninvasive, and cost-effective—this is currently not the case. Imaging tools capable of reliably detecting and localizing aggressive prostate cancer remain limited; as a result, definitive diagnosis still relies on histopathological analysis: examining biopsy tissue under a microscope for signs of cancer. For over two decades, the standard of care has been transrectal ultrasound-guided systematic biopsy. However, this method has been associated with underdiagnosis, overtreatment, and various complications. These limitations underscore the urgent need to explore more accurate and less invasive imaging tools for prostate cancer diagnosis. Recent advancements in magnetic resonance imaging (MRI) and standardized image reporting systems are leading to a significant shift in the diagnostic pathway for prostate cancer. The adoption of multiparametric (mp)MRI and mpMRI-targeted biopsy has become increasingly widespread. Studies show that mpMRI-targeted biopsy improves the detection rate of clinically significant (aggressive) prostate cancer while reducing the identification of indolent (slow-growing, harmless) cases compared to the standard biopsy. In fact, pre-biopsy mpMRI and mpMRI-targeted biopsy are now recommended by guidelines for men with suspected prostate cancer. However, several studies have demonstrated that mpMRI can miss clinically significant prostate cancer and has relatively poor specificity. mpMRI also has practical limitations, including high costs, limited availability, and variability in interpretation. These factors make it less suitable for large-scale screening. As the European Union moves toward image-based prostate cancer screening, interest is growing in more scalable alternatives. Ultrasound, with its low cost, wide availability, and recent technical advances, is gaining attention as a promising option. Among advanced ultrasound modalities, quantitative dynamic contrast-enhanced ultrasound can detect changes in the tumor microvasculature due to angiogenesis, which is associated with prostate cancer growth. In addition to these vascular changes, prostate cancer is associated with increased tissue stiffness. This stiffness can be measured using ultrasound shear-wave elastography—a painless, noninvasive method. Shear-wave elastography uses an ultrasound probe to send sound waves into the body, which push the tissue and create tiny shear waves traveling through the tissue. The ultrasound scanner then measures how fast they move, with shear waves traveling slower in soft tissue and faster in stiff tissue. This allows doctors to assess tissue hardness quantitatively, helping to identify abnormalities such as cancer. Shear-wave elastography has already been implemented in several commercial diagnostic ultrasound systems to assess tissue stiffness, typically using Young's modulus as a measure of elasticity. However, soft biological tissues exhibit both elasticity and viscosity (viscoelastic properties), while most current implementations neglect viscosity, resulting in biased elasticity estimates. Viscosity, in fact, plays an important role in both the propagation and shape of the shear waves. Taking viscosity into account could improve the accuracy of elasticity measurements and may also serve as an additional valuable diagnostic biomarker. This thesis aims to improve elasticity measurements by accounting for viscosity and explores the diagnostic potential of viscoelastic properties in prostate cancer. To support its diagnostic use, this thesis first aims to improve the accuracy of viscoelastic estimation from ultrasound data. A common approach is the phase difference method, which derives a dispersion curve from shear-wave particle motions at two nearby locations to compute local elasticity and viscosity. However, this method is sensitive to noise in the particle-motion measurements. To address this, this thesis proposes two novel methods: the first uses the delayed matrix pencil algorithm to fit noisy shear-wave particle velocity signals; the second applies a local density random walk model to fit displacement signals. Local elasticity and viscosity are then estimated by deriving the dispersion curve from the phase difference between the fitted signals at adjacent locations. Both in-silico and in-vitro experiments demonstrate that these methods are more robust to noise compared to the original phase difference method. Shifting focus from signal processing to clinical application, this thesis introduces a system identification framework for estimating voxel-level elasticity and viscosity in vivo with 3D shear-wave elastography acquisitions, and evaluates their potential role in prostate cancer detection. To capture 3D images, the probe automatically sweeps across the prostate in small steps, collecting 25 cross-sectional images from different angles. Representing the first in-vivo study of viscosity in the prostate, the obtained results in a large patient cohort highlight its promise as a biomarker for prostate cancer localization. Another overlooked aspect of many soft biological tissues is nonlinear elasticity, meaning that their stiffness can change depending on how much they are deformed. Investigating nonlinear elasticity could provide deeper insights into pathologies. This thesis explores for the first time the visco-hyperelastic properties of the prostate, including linear elasticity, viscosity, and nonlinear elasticity. We introduce Volterra kernels to model the local nonlinear shear-wave dynamics, with kernel identification based on shear-wave particle velocities at adjacent locations. We then train a convolutional neural network using features derived from the Volterra kernels and evaluate their performance in silico, in vitro, and in vivo. The findings reveal that Volterra kernels effectively capture visco-hyperelastic properties of prostate tissue, highlighting the potential of nonlinear elasticity as an additional biomarker for prostate cancer detection. In conclusion, this thesis investigates the use of shear-wave elastography to bring prostate cancer diagnosis closer to an accurate, real-time, and cost-effective solution. Specifically, it explores advanced noise reduction techniques and in-vivo estimation of visco-hyperelastic properties in the prostate. The results show that, beyond traditional measures of linear elasticity, both viscosity and nonlinear elasticity hold strong promise as additional biomarkers for localizing prostate cancer by ultrasound imaging. These findings pave the way for more reliable and precise diagnosis using advanced ultrasound techniques. Ultimately, these advancements could reduce the need for unnecessary biopsies and improve patient care, making prostate cancer detection safer, faster, and more accessible.

Proefschrift.