The solid and fluid pressures of tumours are often elevated relative to surrounding tissue. This increased pressure is known to correlate with decreased treatment efficacy and potentially with tumour aggressiveness and therefore, accurate noninvasive estimates of tumour pressure would be of great value. We present a proof-of-concept method to infer the total tumour pressure, that is the sum of the fluid and solid parts, by examining stiffness in the peritumoural tissue with MR elastography and utilising nonlinear biomechanical models. The pressure from the tumour deforms the surrounding tissue leading to changes in stiffness. Understanding and accounting for these biases in stiffness has the potential to enable estimation of total tumour pressure. Simulations are used to validate the method with varying pressure levels, tumour shape, tumour size, and noise levels. Results show excellent matching in low noise cases and still correlate well with higher noise. Percent error remains near or below 10% for higher pressures in all noise level cases. Reconstructed pressures were also calculated from experiments with a catheter balloon embedded in a plastisol phantom at multiple inflation levels. Here the reconstructed pressures generally match the increases in pressure measured during the experiments. Percent errors between average reconstructed and measured pressures at four inflation states are 17.9%, 52%, 23.2%, and 0.9%. Future work will apply this method to in vivo data, potentially providing an important biomarker for cancer diagnosis and treatment.

Solid tumour growth is often associated with the accumulation of mechanical stresses acting on the surrounding host tissue. Due to tissue nonlinearity, the shear modulus of the peri-tumoural region inherits a signature from the tumour expansion which depends on multiple factors, including the soft tissue constitutive behaviour and its stress/strain state. Shear waves used in MR-elastography (MRE) sense the apparent change in shear modulus along their propagation direction, thereby probing the anisotropic stiffness field around the tumour. We developed an analytical framework for a heterogeneous shear modulus distribution using a thick-shelled sphere approximation of the tumour and soft tissue ensemble. A hyperelastic material (plastisol) was identified to validate the proposed theory in a phantom setting. A balloon-catheter connected to a pressure sensor was used to replicate the stress generated from tumour pressure and growth while MRE data were acquired. The shear modulus anisotropy retrieved from the reconstructed elastography data confirmed the analytically predicted patterns at various levels of inflation. An alternative measure, combining the generated deformation and the local wave direction and independent of the reconstruction strategy, was also proposed to correlate the analytical findings with the stretch probed by the waves. Overall, this work demonstrates that MRE in combination with non-linear mechanics, is able to identify the apparent shear modulus variation arising from the strain generated by a growth within tissue, such as an idealised model of tumour. Investigation in real tissue represents the next step to further investigate the implications of endogenous forces in tissue characterisation through MRE.

Thanks to the enhanced T1 dispersion and its linear dependence on temperature over the typical hyperthermia range, T1 mapping represents a serious candidate for temperature mapping at low magnetic fields. Nonetheless, T1 acquisitions are typically long and sensitivity if further reduced at low field regimes. Here we propose a Look-Locker-based temperature mapping sequence that takes advantage of the inherently higher signal gathered using a bSSFP approach. Temperatures spanning from 25 to 50°C could be mapped in doped-water samples with a temporal resolution of 20s/slice and with a precision of 2.6±1.9°C.

Low magnetic field (LF) MRI is currently gaining momentum as a complementary, more flexible and cost-effective approach to MRI diagnosis. However, the impaired signal-to-noise ratio challenges its relevance for clinical applications. Recently, denoising of low SNR images using deep learning techniques has shown promising results for MRI applications. In this study, we assess the denoising performance of residual U-net architecture on different SNR levels of LF MRI data (0.1 T). The model performance has been evaluated on both simulated and acquired LF MRI datasets.