No abstract

No abstract

Purpose. We present a novel background tissue phase removing method, called anatomical phase extraction (APE), and to investigate the accuracy of temperature estimation and capability of reducing background artifacts compared with the conventional referenceless methods. Methods. Susceptibility variance was acquired by subtracting pretreatment baseline images taken at different locations (nine pretreatment baselines are acquired and called φ 1 to φ 9 ). The susceptibility phase data φ S was obtained using the Wiener deconvolution algorithm. The background phase data φ T was isolated by subtracting φ S from the whole phase data. Finally, φ T was subtracted from the whole phase data before applying the referenceless method. As a proof of concept, the proposed APE method was performed on ex vivo pork tenderloin and compared with other two referenceless temperature estimation approaches, including reweighted ℓ 1 referenceless (RW-   ℓ 1) and ℓ 2 referenceless methods. The proposed APE method was performed with four different baselines combination, namely, ( φ 1 , φ 5 , φ 2 , φ 4 ), ( φ 3 , φ 5 , φ 2 , φ 6 ), ( φ 7 , φ 5 , φ 8 , φ 4 ), and ( φ 9 , φ 5 , φ 8 , φ 6 ), and called APE experiment 1 to 4, respectively. The multibaseline method was used as a standard reference. The mean absolute error (MAE) and two-sample t -test analysis in temperature estimation of three regions of interest (ROI) between the multibaseline method and the other three methods, i.e., APE, RW-   ℓ 1, and ℓ 2, were calculated and compared. Results. Our results show that the mean temperature errors of the APE method-experiment 1, APE method-experiment 2, APE method-experiment 3, APE method-experiment 4, and RW-   ℓ 1 and ℓ 2 referenceless method are 1.02°C, 1.04°C, 1.00°C, 1.00°C, 4.75°C, and 13.65°C, respectively. The MAEs of the RW-   ℓ 1 and ℓ 2 referenceless methods were higher than that of APE method. The APE method showed no significant difference ( p > 0.05 ), compared with the multibaseline method. Conclusion. The present work demonstrates the use of the APE method on referenceless MR thermometry to improve the accuracy of temperature estimation during MRI guided high-intensity focused ultrasound for ablation treatment.

No abstract

ABSTRACT:This study describes the application of the method of feedback linearization neural networks, known from neural network computing, to the problem of gradient preemphasis. This approach of preemphasis adjustment does not require an iterative procedure between measurement and adjustment, therefore is essentially instantaneous in its execution. Based on our study, gradient compensation determined by our procedure effectively suppressed eddy current-induced geometric distortion and spatial shift of diffusion-weighted EPI images.