Temporally constrained reconstruction applied to MRI temperature data

Abstract

The monitoring of thermal ablation procedures would benefit from an acceleration in the rate at which MRI temperature maps are acquired. Constrained reconstruction techniques have been shown to be capable of generating high quality images using only a fraction of the k-space data. Here, we present a temporally constrained reconstruction (TCR) algorithm applied to proton resonance frequency shift (PRF) data. The algorithm generates images from undersampled data by iteratively minimizing a cost function. The unique challenges of using an iterative constrained reconstruction technique to generate real-time images were addressed. For a set of eight heating experiments on ex vivo porcine tissue, a maximum reduction factor of 4 was achieved while keeping the root mean square error (RMSE) of the temperature below 0.5 degrees C. For a set of three heating experiments on in vivo canine muscle tissue, the maximum reduction factor achieved was 3 while keeping the temperature RMSE below 1.0 degrees C. At these reduction factors, the TCR algorithm underpredicted the thermal dose by an average of 6% for the ex vivo data and 28% for the in vivo data. Compared with sliding window and low resolution reconstructions, the RMSE of the TCR algorithm was significantly lower (P < 0.05 in all cases).