Improving tumour ADC estimates and elucidating tumour heterogeneity using adaptive Bayesian Markov random field Monte Carlo

Figure 2: Simulated (left) and in vivo (right) ADC point estimates without adaptive smoothing (a and e) and with adaptive smoothing (b and f). ADC uncertainties with adaptive smoothing (c and g) and without adaptive smoothing (d and h). On all maps with adaptive smoothing, the borrowing strength is shown by a border around each pixel; bright lines indicate strong borrowing; dark lines indicate weak borrowing. Point estimates and uncertainty maps are windowed equivalently, illustrating the decrease in uncertainty when adaptive smoothing is implemented. Figure 1: ADC probability density functions for simulated diffusion data (ADC=0.004mm/s) without (black line) and with (grey line) adaptive smoothing. Improving tumour ADC estimates and elucidating tumour heterogeneity using adaptive Bayesian Markov random field Monte Carlo