Search

Proton resonance frequency shift-based magnetic resonance thermometry is a currently used technique for monitoring temperature during targeted thermal therapies. However, in order to provide temperature updates with very short latency times, fast MR acquisition schemes are usually employed, which in turn might lead to noisy temperature measurements. This will, in general, have a direct impact on therapy control and endpoint detection. In this paper, we address this problem through an improved non-local filtering technique applied on the temperature images. Compared with previous non-local filtering methods, the proposed approach considers not only spatial information but also exploits temporal redundancies. The method is fully automatic and designed to improve the precision of the temperature measurements while at the same time maintaining output accuracy. In addition, the implementation was optimized in order to ensure real-time availability of the temperature measurements while having a minimal impact on latency. The method was validated in three complementary experiments: a simulation, an ex-vivo and an in-vivo study. Compared to the original non-local means filter and two other previously employed temperature filtering methods, the proposed approach shows considerable improvement in both accuracy and precision of the filtered data. Together with the low computational demands of the numerical scheme, the proposed filtering technique shows great potential for improving temperature measurements during real-time MR thermometry dedicated to targeted thermal therapies. [ABSTRACT FROM PUBLISHER]

• Deep learning is used for on-line correction of motion artifacts in MR-thermometry • A convolutional neural network (CNN) learns the apparent temperature perturbation • The proposed approach is evaluated on the liver of 12 healthy volunteers • The proposed method is evaluated on a heating experiment performed on a porcine liver • The proposed method is compared to two frequently employed multi-baseline strategies Real-time MR-imaging has been clinically adapted for monitoring thermal therapies since it can provide on-the-fly temperature maps simultaneously with anatomical information. However, proton resonance frequency based thermometry of moving targets remains challenging since temperature artifacts are induced by the respiratory as well as physiological motion. If left uncorrected, these artifacts lead to severe errors in temperature estimates and impair therapy guidance. In this study, we evaluated deep learning for on-line correction of motion related errors in abdominal MR-thermometry. For this, a convolutional neural network (CNN) was designed to learn the apparent temperature perturbation from images acquired during a preparative learning stage prior to hyperthermia. The input of the designed CNN is the most recent magnitude image and no surrogate of motion is needed. During the subsequent hyperthermia procedure, the recent magnitude image is used as an input for the CNN-model in order to generate an on-line correction for the current temperature map. The method's artifact suppression performance was evaluated on 12 free breathing volunteers and was found robust and artifact-free in all examined cases. Furthermore, thermometric precision and accuracy was assessed for in vivo ablation using high intensity focused ultrasound. All calculations involved at the different stages of the proposed workflow were designed to be compatible with the clinical time constraints of a therapeutic procedure. [ABSTRACT FROM AUTHOR]