T2*-weighted MRI as a non-contrast enhanced method for assessment of focal laser ablation zone extent in prostate cancer thermotherapy

Abstract

Objectives:

To evaluate utility of T2*-weighted (T2*W) MRI as a tool for intra-operative identification of ablation zone extent during focal laser ablation (FLA) of prostate cancer (PCa), as compared to the current standard of contrast-enhanced T1-weighted (T1W) MRI.

Methods:

Fourteen patients with biopsy-confirmed low- to intermediate-risk localized PCa received MRI-guided (1.5T) FLA thermotherapy. Following FLA, axial multiple-TE T2*W images, diffusion-weighted images (DWI), and T2-weighted (T2W) images were acquired. Pre- and post-contrast T1W images were also acquired to assess ablation zone (n=14) extent, as reference standard. Apparent diffusion coefficient (ADC) maps and subtracted contrast-enhanced T1W (sceT1W) images were calculated. Ablation zone regions of interest (ROIs) were outlined manually on all ablated slices. The contrast-to-noise ratio (CBR) of the ablation site ROI relative to the untreated contralateral prostate tissue was calculated on T2*W images and ADC maps and compared to that in sceT1W images.

Results:

CBRs in ablation ROIs on T2*W images (TE = 32, 63 ms) did not differ (p = 0.33, 0.25) from those in sceT1W images. Bland Altman plots of ROI size and CBR in ablation sites showed good agreement between T2*W (TE = 32, 63 ms) and sceT1W images, with ROI sizes on T2*W (TE = 63 ms) strongly correlated (r = 0.64, p = 0.013) and within 15% of those in sceT1W images.

Conclusions:

In detected ablation zone ROI size and CBR, non-contrast enhanced T2*W MRI is comparable to contrast-enhanced T1W MRI, presenting as a potential method for intra-procedural monitoring of FLA for PCa.

Introduction

Prostate cancer (PCa) is the most commonly diagnosed cancer and was the third leading cause of cancer death among men in the United States in 2017. [1] Prostate specific antigen (PSA) screening led to an increase in the diagnosis of low-risk and intermediate-risk prostate cancer. [2] The current 5-year survival rate for most men with local or regional prostate cancer is nearly 100%. [3] Traditional treatment of local prostate cancer includes whole-gland radical therapies such as prostatectomy and radiation therapy, or active surveillance. Many indolent PCa-s are thus dramatically over-treated [4, 5], causing unnecessary side effects of radical treatment such as urinary incontinence, erectile dysfunction, and bowel toxicity. [6, 7] On the other hand, active surveillance may cause psychosocial and financial burdens for the enrolled patients. [8]

In order to achieve the oncological benefit of active treatment yet avoid the genito-urinary and rectal side effects, in recent years focal therapy has been introduced as a treatment for localized PCa. MRI-guided focal laser ablation (FLA) is a novel technique for focal therapy of prostate cancer and involves thermal destruction of PCa tissue by laser while reducing the risk of healthy adjacent normal tissue damage relative to conventional treatments. The structure, water content and perfusion of the tissue determines its optical and thermal properties, which influence the ability to convert the laser light into heat in tissue, [9] and prostate tissue is well-suited for FLA because of its high optical absorption rate and relatively low vascularity. [9] The laser produces accurate, predictable, and reproducible ablation zones and causes minimal changes to adjacent normal prostate tissue, therefore avoiding, to a large extent, treatment-related toxicity. [10] In addition, MR imaging can produce good soft tissue contrast and anatomical detail to guide treatment planning and targeting, making MR-guided FLA superior to other ablation techniques. Some previous trials reported feasibility and favorable safety and oncologic outcomes of MR-guided FLA, [9-14] and MRI can also be used to track treatment response. [15]

For successful focal therapy, it is important to accurately assess the ablation region size and to validate that the targeted area is ablated, while minimizing damage to surrounding tissue. Presently, gadolinium-based contrast enhanced T1-weighted (T1W) imaging is used to assess the extent of coagulated necrosis caused by thermal therapy of tissues, such as in uterine fibroid ablation, [16] laser photocoagulation of breast [17] and brain [18], ultrasound ablation of prostate cancer, [19] as well as in FLA of prostate cancer, [20, 21] and good correlation of ablation area between MRI and histology was demonstrated. However, gadolinium-based contrast agents can be retained in the ablated tissue for several hours, which not only makes repeated evaluation ineffective, but could potentially result in toxicity if the stability of the contrast chelate is compromised by sharply increased tissue temperatures. Therefore, the practical use of the contrast-enhanced sequence is limited to one-time at the final assessment.

A non-contrast enhanced MRI sequence would be preferred as it could be repeated many times during FLA, to assess the ablation zone and plan the ablation intra-procedurally, and this is an area of ongoing effort. MR thermometry based on the water resonance frequency shift is wildly used to monitor in real time the thermal damage of MR-guided thermotherapy, as we used in the FLA procedure. It is reported that the lesion sizes measured by real-time MR thermometry during procedure of radiofrequency ablation of myocardium correlated well with the lesion size after ablation on contrast-enhanced T1W images and macroscopic measurements[22]. But Boomer et.al[23]. reported that the ablation volumes on MRI-temperature maps during MRI-guided focal laser ablation for prostate cancer were always larger than the ablation zones in the T1-weighted contrast-enhanced MR images after ablation and the necrotic volumes in the histopathology specimen, the reason of which could be the tissue heterogeneity within the prostate, thermal conductivity, change caused by temperature alteration, spatial resolution of the sequence and patient motion [23]. Contrast on diffusion-weighted imaging (DWI) has been shown to correlate with non-viable areas on histopathological analysis after thermal ablation of prostate cancer. [24] Staruch et al reported that the acute thermal damage in muscle caused by high-intensity focal ultrasound (HIFU) using magnetization-prepared 3D T2-weighted (T2W) imaging correlated well with contrast-enhanced T1W imaging. [25] Lam et al. used a multi-echo gradient echo (MEGE) MRI to monitor near-field HIFU heating. [26] However, to date, no non-contrast enhanced MR imaging technique has been established for intra-procedural assessment in prostate FLA.

A previous study assessed thermal damage following FLA treatment of PCa patients using T2* maps and water peak height images calculated from T2*-weighted (T2*W) images acquired with a MEGE sequence.[27] The study showed that the contrast-to-noise ratio (CBR) of ablated regions in T2* map and water peak height images were not significantly different from that in contrast-enhanced T1W images. However, the acquisition and post-processing of MEGE images carries a time penalty, and using individual T2*W images for FLA monitoring would reduce acquisition times and/or allow higher resolution imaging, and eliminate some post-processing steps.

The purpose of this study is to evaluate the feasibility of using single-TE T2*W MR imaging for identification of acute ablation zone extent following FLA of PCa, as compared to ADC mapping and contrast-enhanced T1W MRI.

Materials and Methods

Patients

This prospective study was approved by our institutional review board. Written informed consent was obtained from all subjects. We recruited fourteen men with biopsy-confirmed low- to intermediate- risk localized prostate cancers as an add-on to our Phase II clinical trial for MR-guided FLA. A transrectal ultrasonography-guided biopsy with a minimum of 12 biopsy cores were performed. The inclusion criteria were: prostate specific antigen (PSA) less than 15 ng/ml or PSA density less than 0.15 ng/ml²; clinical stage T1c-T2a, Gleason score 7 or less in 25% or less of biopsies and MR imaging with 1 or 2 lesions concordant with biopsy detected cancer. The mean age of patient was 61 years (range 51-71 years old, median 64 years old), and mean PSA was 4.2 ng/ml (range 0.9–8.9 ng/ml, median 4.6 ng/ml). PCa was diagnosed from previous MRI screening scans and biopsy about 1-4 months before the FLA in our institution. Gleason score identified on biopsy was 6 (3+3) in all 14 patients, and the lesions of all patients were located in the peripheral zone.

MRI scan

The MR imaging and FLA were performed under monitored conscious sedation. Patients were placed in the supine, head first position on a 1.5 T MR scanner (Achieva 1.5T-TX, Philips Healthcare) equipped with a Quasar Dual gradient system (40 mT/m; 200 T/m/s) and local, subject-adaptive RF shimming using dual-source RF transmission (MultiTransmit). An 8-channel cardiac phased array receiver coil and a single channel endo-rectal coil (Medrad eCoil) were used in all 14 patients.

On the day of FLA, high resolution T2W images were acquired. An experienced radiologist linked the screening MRI and biopsy results with the intra-procedural high resolution T2W images for fiducial localization and laser fiber placement, after which the ablation procedure proceeded.

After the ablation procedure, axial 2D MEGE images with free breathing (TR = 1721 ms, first TE = 3.41 ms, subsequent echo spacing = 3.2 ms, number of echoes = 20, flip angle = 25°, FOV = 304×92×380 mm³, in-plane resolution = 2.3×2.3×3.8 mm³, reconstructed voxel size = 1.04×1.05×3.8 mm³, number of slices = 23, no gap between slices, averages = 1, scan time = 115 s). From these, T2*W images obtained at TE = 3.4 ms, 32 ms, and 63 ms (T2*W_3.4, T2*W₃₂, and T2*W₆₃) were obtained for further analysis.

Additionally, axial DWI with spin echo-echo planar imaging sequence (SE-EPI DWI, TR = 7834 ms, TE = 74 ms, flip angle = 90°, b-values = 0, 600 s/mm², FOV = 400×400 mm², in-plane resolution =2.0×2.0×3.0 mm³, number of slices = 30, averages = 3, scan time = 186~210 s) was acquired. Adynamic contrast-enhanced MR sequence was scanned for 5-8 minutes after intravenous administration of contrast agent (0.1 mmol/kg, gadobenate dimeglumine, Multihance, Bracco Diagnostic Inc.). High resolution T1W images were acquired using the 3D eTHRIVE sequence (TR/TE = 3.9/1.8 ms, FOV = 308×240×380 mm³, flip angle = 10°, in-plane resolution = 1.6×2×4 mm³, reconstructed voxel = 1.0×1.0×2.0 mm³, number of slices = 100, averages = 1, scan time = 18 s) before intravenous contrast agent injection and after the dynamic contrast-enhanced sequence. Subtracted contrast-enhanced T1W (sceT1W) images were generated by subtracting pre-contrast T1W images from post-contrast T1W images on a workstation. Apparent diffusion coefficient (ADC) maps were calculated from DWI data using a mono-exponential signal decay model on the MR console.

FLA Procedure

FLA was performed using the Visualase system (Medtronic). The titanium trocars and guide catheters were placed transperineally using an MR-compatible transperineal template fixed to the endorectal coil. A laser applicator was then placed through the catheter and inserted into the satisfactory location in the target lesion. Regions of thermal damage were produced by a 15 W laser beam. Each individual laser ablation lasted 90-120 seconds. Repeated ablations were necessary for complete treatment of the lesions, for an average total ablation time of 4.3 min. [10, 11, 28]. The overall duration of the treatment procedure was 2.5-4 hours.

Tissue temperature was monitored by MRI during FLA based on the proton resonance frequency offset method. [29, 30] A temperature-sensitive fast radiofrequency-spoiled gradient-recalled echo T1W sequence (TR/TE = 40/20 ms, in-plane resolution = 1.2×2.5×5 mm³; reconstructed voxel = 1.2×1.2×5.0 mm³, flip angle = 20°; dynamic time= 5 seconds) was run repeatedly during FLA and the images were processed on the Visualase system to produce real-time temperature maps. Tissue damage estimates from the software based on Arrhenius evaluation of temperature history were superimposed on the magnitude anatomic MR images and displayed to aid in treatment decision-making. Target ablation temperature was a minimum of 60°C. When important functional structures (eg cavernous nerves and urethra) were not adjacent, the intent was ablation extending more than 5 mm beyond the lesion. The urologist and radiologist subjectively compared the non-enhancing areas on immediate postcontrast images with the temperature and thermal damage maps to verify the targeted lesion was completely ablated.

Imaging Data Analysis

Regions of interest (ROIs) for the resulting ablation areas were identified and manually outlined by an experienced radiologist (10 years of prostate MRI experience) on T2*W images and sceT1W images, who was blinded to pre-procedure scans and intraprocedural imaging. The ROIs from sceT1W images were propagated to ADC maps. The slice with the largest ablated ROI in each case was used for quantitative analysis, and the ROI sizes were calculated there. CBR of ablated ROIs was calculated on T2*W images, ADC maps, and sceT1W images: the difference in signal intensity (SI) between the ablated ROI and that in an ROI in untreated peripheral zone prostate tissue contralateral to the ablation site was normalized to the standard deviation of the signal intensity in the untreated prostate ROI (Eq. 1).

$$CBR=\frac{\mid mean(S{I}_{ablated})-mean(S{I}_{untreated})\mid }{std(S{I}_{untreated})}$$ (1)

Untreated prostate tissue was used for reference since the pelvic muscles were not well visualized in the vicinity of the ablated regions. A ROI with the similar size and shape as the ablated region ROI was placed symatrically in the contralateral untreated prostate tissue as reference(figure 1-f. and figure 2-f).

Figure 1:

A representative example with biopsy-proved prostate cancer in right peripheral zone of prostate. (a) Pre-procedure T2-weighted axial fast spin-echo image shows focal hypointense area within right peripheral zone posteriorly(circle), which is indicative of cancer. (b) Pre-procedure Apparent diffusion coefficient map through same level shows a focal area of low signal intensity (arrow). The ablated region of prostate cancer post-FLA in the right peripheral zone was shown on sceT1W image (c.), T2*W_3.4(d.), T2*W₃₂(e), T2*W₆₃ image (f), and ADC map (g.). The CBR values in a., b., and c. were 3.14, 0.10, 2.25 3.47 and 0.55 respectively. The ROI of ablated region was manually drawn by a radiologist, and another ROI (red) of approximately the same size and shape was placed symmetrically on the contralateral untreated prostate tissue(f.), which was used for reference.

Figure 2.

Another representative example with a biopsy confirmed prostate cancer in the right peripheral zone. On the Pre-procedure T2-weighted image (a.) and apparent diffusion coefficient map (b.), the prostate cancer shows low signal intensity in the right posterior peripheral zone. The ablated region of prostate cancer after FLA was shown on sceT1W image (c.), T2*W_3.4(d.), T2*W₃₂(e.), T2*W₆₃ image (f.), and ADC map (g.). The CBR values in c., d., e., f. and g. were 2.95, 1.91,1.54, 2.08 and 0.49 respectively. Example of reference ROI (red) was shown in T2*W₆₃ image (f.).

Statistical analysis

The non-parametric Friedman’s test with post-hoc two-tailed Wilcoxon signed-rank test was performed to determine whether there was a statistically significant difference between CBRs derived from T2*W images, ADC maps, and sceT1W images, and between ROI sizes derived from T2*W₃₂ and T2*W₆₃ images and sceT1W images. Bland-Altman analysis was performed to evaluate the agreement between ROI sizes and CBRs of T2*W₃₂ and T2*W₆₃ images vs sceT1W images, as well as relative ROI difference (%) with regard to sceT1W images of T2*W₃₂ and T2*W₆₃ images. The correlation between ROI sizes measured in T2* W₆₃ and sceT1W images was evaluated using the Pearson’s coefficient of correlation. A p-value of less than 0.05 was considered statistically significant.

Results:

Representative examples of ablation area outlined on sceT1W images, T2*W images, and ADC maps are shown in Figures 1 and 2. The CBR values in ablation ROI in T2*W_3.4, T2*W₃₂, and T2*W₆₃ images, ADC maps, and sceT1W images, as well as ablation ROI sizes in T2*W₃₂ and T2*W₆₃ images and in sceT1W images, are given in Table 1, with p-values from pairwise comparison to sceT1W values. The distributions of these values are shown in Figures 3a and 3b. CBR values in ablation ROIs in T2*W₃₂ and T2*W₆₃ images were not statistically significantly different from those in sceT1W images and ablation ROI sizes were compared for only these sets of images. CBR values in ablation ROIs in T2*W_3.4 images and in ADC maps were statistically significantly lower than those in sceT1W images, thus ablation ROI size comparisons were not made.

Table 1:

Ablation zone ROI CBRs and sizes for T2*W images, ADC maps, and sceT1W images

	T2*W3.4	T2*W32	T2*W63	ADC	sceT1W
values
ROI CBR	1.7±1.2	2.9±1.8	2.7±1.3	0.9±0.6	3.3±1.0
ROI size [cm2]		21.4±1.5	24.1±1.7		28.3±2.0
Relative ROI
size[%]		76.2±14.2	86.5±16.7
p-values
ROI CBR	0.06	0.33	0.25	0.01
ROI size		0.001	0.016

Figure 3.

(a.) Boxplots of CBRs of ablated region ROIs calculated from T2*W_3.4, T2*W₃₂, and T2*W₆₃ images, ADC maps, and sceT1W images are shown. (b.) Boxplots of ROI sizes for T2*W₃₂, T2*W₆₃, and sceT1W images are shown.

Bland-Altman plots of ablation ROI CBR values and sizes showed good agreement of T2*W₃₂ and T2*W₆₃ images vs. sceT1W images (Figure 4). Comparing ROI sizes in T2*W₆₃ and sceT1W images, we found good correlation (Pearson’s r = 0.64, p=0.013), with ROI sizes in T2*W₃₂ and T2*W₆₃ images being statistically significantly lower by approximately 24% and 15% (p = 0.001, 0.016), respectively.

Figure 4.

Bland–Altman plots show good agreement between ablation zone ROI CBR values on T2*W₃₂ and T2*W₆₃ images vs sceT1W images (a, b). Bland–Altman plots show good agreement between ablation zone ROI sizes on T2*W₃₂ and T2*W₆₃ images vs sceT1W images (c, d), as well as Relative ROI size difference (%) with regard to sceT1W images on T2*W₃₂ and T2*W₆₃ images (e, f).

Discussion

We examined the feasibility of using non-contrast T2*W MRI, without additional post-processing, for continuous monitoring of FLA therapy for PCa. The CBR values of T2*W₃₂ and T2*W₆₃ images were comparable to those obtained from the current-standard sceT1W imaging, and the ablated region sizes in T2*W₆₃ images were well-correlated with and within 15% of those observed in sceT1W images. Thus, T2*W MRI could be useful for intra-operative monitoring of the ablation zone extent, with the post-operative contrast-enhanced imaging used for final assessment. It is also likely that with increased spatial resolution, e.g., at a higher field, the discrepancy of ROI sizes between the two sequences would be reduced.

Sizes of ablated region in T2*W₃₂ and T2*W₆₃ images showed approximately 24% and 15% underestimation of sceT1W ablation size respectively. The contrast between the ablated lesion region and the untreated prostate tissue on the contrast-enhanced T1W images was much higher compared with T2* images. This difference of contrast may lead to an increase of ablation area measured on postcontrast T1W compared with T2* images.

Based on our results, echo times beyond 63 ms could potentially result in even better agreement between T2*W and the current standard sceT1W images in terms of ablation zone size. T2*W_3.4 images and ADC maps showed inferior CBRs and are not good candidates for FLA monitoring. Our results show that T2* is a useful contrast in this context, and that the increase in T2* weighting with higher TE values increases the utility of the sequence, in agreement with an earlier study[27]. Ablation sites had lower signal compared to untreated prostate on all T2* images. The hypo-intensity on T2* images is likely caused by the dehydration, carbonization, and coagulative necrosis [25, 31], which induces local susceptibility differences, leading to a decreased T2* value, which was observed in an earlier study [27].

In an earlier study by Staruch [25], an innovative magnetization-prepared 3D T2-weighted sequence, SHINKEI [32] was tested in muscle and demonstrated an advantage in showing ablation lesions as hyper-intense. However, it has a rather long acquisition time (~11 minutes in the referenced study), and its contrast is dependent on the accumulation of edema in the treated region. This makes SHINKEI impractical for the purpose of intra-procedural monitoring of FLA in prostate. In contrast, T2*W imaging, proposed here, can be both fast and implemented immediately after an ablation session.

It has been reported previously that DWI and ADC mapping could be used to monitor prostate thermal therapy. In a study of an animal model [33], in which b values of 30 and 380 s/mm² were used, the ADC value of prostate tissue was reduced immediately after thermal treatment. Another study of HIFU ablation of uterine fibroids [34] showed that ADC value decreased post-treatment when calculated using low b-values (0 and 200 s/mm²), increased when calculated using high b values only (400, 600 and 800 s/mm²), and was unchanged when b values of 0 and 800 s/mm² were used – in general agreement with our results. This points to a combination of reduction in perfusion via occlusion of small blood vessels by thermal coagulation and a simultaneous increase in diffusion via destruction of cell membrane [35] . This complex dependence of changes in ADC on the specific selection of b values, as well as the low spatial resolution, diminish potential utility of DWI in thermotherapy monitoring.

There were some limitations in this study. Most importantly, as this study focused on focal therapy, prostatectomies were not performed, and the actual ablated region sizes could not be confirmed on histopathology. The good correlation of ablation ROI sizes in T2*W₆₃ images with those measured in the current standard sceT1W images is encouraging, though. Further, the ablation zone ROIs were drawn by hand and it is possible that the differences in image contrast and resolution led to a systematic under- or over-estimation of the zone between image sets (e.g., T2*W₆₃ vs sceT1W). This could potentially be alleviated by using automatic segmentation algorithms, and acquisition of single-TE T2*W images would allow for higher spatial resolution and more accurate delineation of ablation zones, relative to the current MEGE sequence. In addition, we used a 1.5-T MR unit for procedure guidance. Using a 3.0-T unit for ablation guidance may increase the signal-to-noise ration and spatial resolution, therefore, smaller lesions could be visualized and treated by FLA. However, the optimal TE for T2*W imaging is still under evaluation. Finally, the utility of DWI in this study could have been limited by the specific b value selection. Thus, our results showing very low CBR should not discourage further exploration of ADC maps for ablation zone delineation, using higher and/or lower b values in future imaging protocols.

In conclusion, we show that non-contrast T2*W MR images acquired at a long TE visualize the extent of post-procedure ablation zone comparably to the current standard contrast-enhanced T1W MRI, in terms of ROI contrast and size. This demonstrates feasibility of using T2* MRI as a plausible method for thermal ablation zone delineation. The ability of T2*W MRI to visualize the ablation zone without use of contrast agents could allow intra-operative monitoring of FLA for PCa, potentially improving treatment outcomes.

Key points:

• T2*-weighted MR images with long TE visualize post-procedure focal laser ablation zone comparably to the contrast-enhanced T1-weighted MRI.

• T2*-weighted MRI could be used as a plausible method for repeated intra-operative monitoring of thermal ablation zone in prostate cancer, avoiding potential toxicity due to heating of contrast agent.

Abbreviations:

ADC

Apparent diffusion coefficient

CBR

Contrast-to-background ratio

DWI

Diffusion weighted imaging

FLA

Focal laser ablation

Pca

Prostate cancer

ROI

Region of interest

sceT1W

Subtracted contrast-enhanced T1-weighted

T1W

T1-weighted

T2*W

T2*-weighted