Introduction

Microwave ablation is a novel way of treating liver cancer lesions¹. Usually these procedures are performed under CT guidance where the interventional radiologist has no thermal feedback regarding heating in the tumor. Proton Resonance Frequency (PRF) thermometry is a widely used MRI based technique to monitor changes in tissue temperature in response to thermal therapy². However, motion related anatomical misregistration and background phase variation are two of the key factors limiting clinical application of PRF thermometry to guide microwave ablation procedures in the liver³. In this abstract, we present a novel acquisition and processing pipeline for liver thermometry to overcome these challenges. We propose the use of segmented 3D Echo Planar Imaging (EPI) to permit 3D deformable registration, and to improve signal to noise ratio (SNR) and spatial coverage. Principle component analysis (PCA) based baseline correction is used to remove baseline phase variation.

Methods

Data Acquisition: All human imaging protocols were approved by local Institutional Review Board. One healthy volunteer was imaged on a 3T scanner (MAGNETOM Prisma fit, Siemens Healthcare, Erlangen, Germany) using a prototype 3D segmented EPI sequence⁴ modified to incorporate interactive pause capability. The following parameters were used: TR 24ms, TE 5.1ms, fat saturation, 2x3x3mm³ spatial resolution, 192x105x10 matrix with 20% slice oversampling, EPI factor 9, in-plane acceleration R = 2, 100 repetitions, 1.75 seconds per 3D volume. Interactive pause functionality was used to divide the acquisition into ~16-20s breath-holds.

Preprocessing: Each image was registered to the first image in the series using a deformable 3D registration algorithm⁵. First image of each breath-hold was discarded to avoid transient magnetization effects. First 40 images of the resultant series were chosen as baseline images (referred to as baseline series). The remaining images (referred to as therm series) were used to compute temperature difference maps.

Background phase removal and ΔT estimation: We utilized a PCA based algorithm to remove motion related phase changes⁶ : Complex mean of the baseline images B_Ave was obtained. PCA was performed over a liver mask on the series consisting of unwrapped complex phase difference between baseline images and B_Ave. First 20 of the resultant eigen-images were retained as basis images for correcting the therm series. Complex phase difference was computed between each therm image and the corresponding previous therm image. B_Ave was used as the image before the first therm image. Baseline correction was performed by removing the projection of each unwrapped difference image onto the subspace spanned by the basis images. Temporal cumulative sum of the resultant series was obtained.

For comparison, we also performed baseline correction with the Dictionary-based approach, where the best matching baseline image is looked up for each therm image⁷. For both approaches, average phase over the mask was removed after baseline correction to eliminate global phase drift. Resultant phase difference images were scaled by -1/(γB₀xTE×0.01ppm/°C) to estimate temperature difference relative to the first acquisition. Since the expected temperature change was 0 in absence of external heating, we used temporal mean (μ_T) and standard deviation (σ_T) of the estimated ΔT series to assess estimation bias and variability.

Results

3D deformable registration adequately corrects inter-breathhold motion, as evidenced by a reduction in temporal standard deviation of magnitude images (Figure 1). Thermometry maps obtained using PCA based approach exhibit excellent precision and accuracy (Figure 2). |μ_T| and σ_T for PRF measurements corrected using the dictionary-based are greater than those corrected with the PCA-based approach (Figure 2, 3 and 4).

Discussion

Our results suggest that PRF measurements acquired and processed using the proposed pipeline exhibit excellent precision and accuracy. |μ_T| and σ_T are much smaller in most of the liver than the threshold for irreversible tissue damage³, suggesting the feasibility of PRF based guidance for liver ablation. Consistent with results reported by Majeed et al⁶, the PCA based approach outperforms the existing dictionary-based approach.

To our knowledge, all liver thermometry studies to date have utilized 2D acquisition schemes. Although these studies rely on navigators, respiratory triggering and breath-holds to reduce the impact of respiratory motion, residual inter-scan motion cannot be avoided. Limited spatial coverage in slice direction makes 3D registration unfeasible, resulting in compromised anatomical precision. The proposed segmented EPI based approach overcomes this challenge while keeping the frame rate within reasonable limits. Additionally, 3D acquisition results in higher SNR for a given resolution and spatial coverage.

We have chosen to use TE = 5.1ms, which is much smaller than the T2* of the normal liver at 3T (14.5±6.7ms)⁸. Although TE = T2* is the optimal choice of TE for PRF thermometry⁸, TE values smaller than T2* of the normal tissue may be required in practice to avoid intravoxel dephasing due to B₀ inhomogeneity induced by the ablation device. 3D acquisition may prove crucial to counteract the resulting loss of temperature sensitivity and SNR.

The interactive pause functionality described in this study enables multi-breathhold acquisitions without the need to stop the scan. Additionally, it can be used to pause the acquisition during microwave-based ablation to avoid unnecessary acquisition of noise corrupted images.

Future work will focus on evaluation on a larger sample, further protocol optimization and comparison with additional existing approaches.

Acknowledgements

No acknowledgement found.