Temperature-Sensitive Frozen Tissue Imaging for Cryoablation Monitoring Using STIR-UTE MRI

Abstract

Purpose.

To develop a method to delineate the lethally frozen tissue region (Temp<−40°C) arising from interventional cryoablation procedures using a short tau inversion-recovery ultrashort echo-time (STIR-UTE) MRI sequence. This method could serve as an intra-procedural validation of the completion of tumor ablation, reducing the number of local recurrences after cryoablation procedures.

Materials and Methods.

The method relies on the short T₁ and T₂* relaxation times of frozen soft-tissue. PETRA, a 3D UTE sequence with TE = 70 μs was optimized with STIR to null tissues with T₁ ~ 271 ms, the threshold T₁. Since the T₁ relaxation time of frozen tissue in the temperature range of −40°C < Temp < −8°C is shorter than the threshold T₁ at the 3 Tesla magnetic-field, tissues in this range should appear hyperintense. The sequence was evaluated in ex vivo frozen tissue, where image intensity and actual tissue temperatures, measured by thermocouples, were correlated. Thereafter, the sequence was evaluated clinically in 12 MR-guided prostate-cancer cryoablations, where MR-compatible cryoprobes were used to destroy cancerous tissue and preserve surrounding normal tissue.

Results.

The ex vivo experiment using a bovine muscle demonstrated that STIR-UTE images showed regions approximately between −40°C and −8°C as hyperintense, with tissues at lower and higher temperatures appearing dark, making it possible to identify the region likely to be above the lethal temperature inside the frozen tissue. In the clinical cases, the STIR-UTE images showed a dark volume centered on the cryoprobe shaft, V_inner, where the temperature is likely below −40°C, surrounded by a doughnut-shaped hyperintense volume, where the temperature is likely between −40°C and −8°C. The hyperintense region was itself surrounded by a dark volume, where the temperature is likely above −8°C, permitting calculation of V_outer . The STIR-UTE frozen-tissue volumes, V_inner and V_outer, appeared significantly smaller than signal voids on turbo spin echo images (p<1.0⨉10⁻⁶), which are currently used to quantify the frozen-tissue volume (“the iceball”). The ratios of the V_inner and V_outer volumes to the iceball were 0.92 ± 0.08 and 0.29 ± 0.07, respectively. In a single post-ablation follow-up case, a strong correlation was seen between V_inner and the necrotic volume.

Conclusion.

STIR-UTE MRI successfully delineated the area approximately between −40°C and −8°C isotherms in the frozen tissue, demonstrating its potential to monitor the lethal ablation volume during MR-guided cryoablation.

Introduction

Prostate cancer (PCa) is the most common malignant tumor and the second most common malignancy leading to death for men in the United States ¹. While radical treatments (i.e. prostatectomy and radiotherapy) have been commonly performed, their excessive use has raised concerns over overtreatment and unnecessary risk of complications (e.g. incontinence and impotence) ^2-5. The recently growing proportion of low- and intermediate-risk disease detected in newly diagnosed PCa ⁶ have increased these concerns.

Prostate focal cryoablation has been performed as a minimally-invasive option for primary treatment of localized, low- and intermediate risk PCa and salvage treatment after primary radiation therapy ^7-12. Like other ablative modalities used for focal treatments, such as high-intensity focused ultrasound (HIFU) ^13-15 and laser interstitial thermotherapy ^16-19, cryoablation can target a localized diseased area and leave the rest of the organ intact, which avoids surgical complications such as incontinence and impotence ^7,9. While there is no definitive conclusion as to which modality leads to the best clinical outcome, cryoablation provides several technical advantages, including the good visibility of the ablation volume and a proven effectiveness in salvage treatment. In particular, intraprocedural MRI is known to provide excellent visualization of the tumor ²⁰. It also visualizes the volume frozen to a temperature (Temp) which is below 0°C, which is referred to as the “iceball”, making it an attractive tool for cryoablation monitoring. Conventional MRI spin-lattice relaxation time (T₁-) or spin-spin relaxation time (T₂-) weighted sequences, which have a time-to-echo (TE) ≥1ms, both show the entire iceball as a signal void, due to the short spin dephasing (T₂*) relaxation time of tissues below 0°C, where T2*<<1ms.

However, the iceball volume may not precisely represent the area of lethal ablation, because studies have shown that temperatures in the −20°C to −40°C range are required to achieve total necrosis in cancerous tissues ²¹. This led to the question of whether conventional MRI may be overestimating the actual ablation zone. To delineate the lethal ablation zone, MR thermometry of frozen tissue using ultrashort echo time (UTE) MRI sequences was proposed ^22-25. Those techniques estimate temperature based on quantifying the tissue T₂*, which is temperature dependent, or using the normalized UTE signal intensity, which is proportional to T₂*. Clinical application of these techniques is, however, not always straightforward, as they require calculations of T₂* based on multiple images, obtained at different TEs, which may be difficult to perform in regions susceptible to physiological motion.

In this study, we propose a novel approach to predicting the lethal ablation zone using a short tau inversion recovery ultrashort echo-time (STIR-UTE) MRI scan with a single TE, which provides temperature-sensitive contrast within the iceball. The UTE sequence employs a Pointwise Encoding Time with Radial Acquisition (PETRA) readout ²⁶. The new approach is first tested in an ex vivo tissue experiment, and then evaluated in vivo during focal prostate cancer cryoablations

Materials and Methods

Temperature-Sensitive MRI Based on STIR-UTE PETRA

Our approach relies on studies that have shown that T₁ relaxation time of frozen tissue is strongly dependent on temperature and the Larmor frequency ^27,28. The T₁ drops sharply as the temperature drops below −8°C, reaches a minimum at approximately −20°C to −40°C, depending on the magnetic-field strength, and then gradually grows as temperatures further decrease. As a result, if the tissue T₁s at the upper and lower temperature limits are known, the volume within this temperature range can be readily delineated with a STIR sequence that selectively suppresses signals with longer T₁s. The boundary temperatures are determined by the selected inversion time (TI). In addition, since the echo time (TE) of tissues in the iceball are extremely short (typically < 200 μs), the STIR sequence must be combined with a UTE readout in order to observe these tissue components.

The relationship between temperature and T₁ relaxation time of frozen soft-tissue (rat muscle) at fields between 0.5 and 3 Tesla (T) is shown in Figure 1, based on a validated model provided by Fung and McGaughy ²⁸. Assuming that the lower boundary (critical) temperature is −40°C, the T₁ relaxation time at the critical temperature is estimated to be T_{1, −40°C} = 271 ms at 3 Tesla. Therefore, if signals from species with relaxation times around 271 ms are either nulled or suppressed, only the area where −40°C < Temp < −8 °C should appear hyperintense.

Fig. 1.

Estimated T1 relaxation time in soft-tissue (muscle) in frozen and unfrozen conditions at 0.5 Tesla, 1.0 Tesla, 1.5 Tesla, and 3.0 Tesla based on equations and parameters provided in reference ²⁸. We computed the T1 relaxation times for these field strengths since the original work did not include a 3 Tesla plot. Arrows and the horizontal dotted lines indicate the lower and upper temperatures where the T1 at 3T is 271 ms. Note that the freezing temperature in soft-tissue is depressed by 5 to 10 degrees, relative to bulk water, due to surface tension.

In this study, a 3D isotropic STIR PETRA sequence ²⁶ was used to obtain the UTE images, with a single STIR pulse preceding several (N>>1) radial readouts (spokes). To keep the acquisition time short enough (~90 seconds/volume) to continuously track the growth of the iceball during the cryoablation procedure, a minimal repetition time per spoke, TR, was selected, along with a relatively small matrix size. When a STIR 3D UTE sequence with a multiple-spoke acquisition, like PETRA, is used with a relatively low flip angle, the inversion time required to null areas with a specific T₁ is given by ²⁹:

$$T_I=-T_1\text{ln}\left(\frac{1+e^{-T_{R0}∕T_1}}{2}\right)$$ (1)

Where T_R0 =T_I + N T_R+T_Rel is the total time per pass or the interval between inversion pulses, and T_Rel is the relaxation time allowed after readout to allow for magnetization recovery. When T_R0 = 200 ms is selected, the inversion time (T_I) to null the frozen tissue at the (−40°C) critical temperature is approximately 80 ms. Tissues with longer T₁s (~1200 ms), such as unfrozen soft-tissue, are suppressed as well, because the short T_R0 used (T_R0 << T_{1soft-tissue)}) does not allow for sufficient relaxation time between the inversion pulses.

The acquisition time of PETRA was TA=84 s per 3D volume using the following parameters: TI/TR0/TR/TE = 80/200/1.6/0.07 ms; number of spokes = 15000, number of spokes after each inversion pulse (N) = 40, matrix size = 96³, voxel size = 2.0³ mm³, FOV = 240³ mm³, and Flip angle = 10°.

Pre-Clinical Evaluation

We performed two sets of experiments using a 3 Tesla MRI scanner (MAGNETOM Verio, Siemens Healthineers, Erlangen, Germany) to test (1) if the STIR-UTE sequence nulls the signal from areas with T₁ ~ T_{1, −40°C} and T₁ ~ T_{1, −8°C} and suppresses the signals from areas with larger T₁ and (2) if the STIR-UTE sequence provides a temperature-sensitive contrast in a frozen tissue.

T₁ sensitivity in a phantom

We validated the T₁ sensitivity of the STIR-UTE sequence by acquiring images of T₁-mapping phantoms with known T₁ relaxation times. Nine T₁-mapping phantoms were constructed in separate cylindrical vials by mixing agarose gel with varying nickel chloride concentrations ³⁰. The vials were mounted parallel to each other and collectively imaged.

The reference T₁ relaxation time of the gel within each vial was quantified using an STIR Turbo Spin Echo (STIR-TSE) sequence, repeated with multiple TIs (TR/TE = 3000/15ms, TI = 24, 50, 80, 108, 146, 198, 268, 363, 490, 663, 896, 1640, 2200, 2800 ms, flip angle = 180°, echo train length = 9, matrix = 320⨉260, pixel size = 0.5² mm², slice thickness = 2mm, FOV=160⨉130mm). The mean signal intensity in a region-of-interest within each vial, at each TI, was recorded using 3D Slicer software ³¹. The T₁ relaxation time for each T₁-mapping phantom was then obtained by fitting the data to the standard model for T₁ relaxation.

The T₁-mapping phantoms were then scanned using the 3D STIR-UTE PETRA sequence with the parameters described previously. The mean signal intensity within each vial was plotted against the T₁ relaxation time in order to verify that the signals from the phantoms with T₁ ~ T_1,−40°C and T₁ ~ T_1,−8°C (i.e. 271 ms) were nulled and the signals from areas with larger T₁ were suppressed. As the STIR-UTE PETRA images presented with spatial intensity nonuniformity ³², due to “streaking” artifacts (such as radial lines and cubic matrices) not corrected by image reconstruction, which made it difficult to compare pixel intensities in different regions of the image, an intensity bias field was estimated using the 3D Slicer N4ITK MRI bias correction ³³ with a grid size of 5⨉5⨉5. Then the intensity nonuniformity was then corrected before measuring the signal intensity.

Temperature sensitivity in a frozen tissue sample

An ex vivo experiment was performed to test if the STIR-UTE image could visualize the isotherm for the −40°C critical temperature in a frozen tissue. Slices of swine muscles (approximately 1 cm thickness) were stacked in layers within a Styrofoam container (Figure 2). To create a temperature gradient in the sample, a block of dry-ice (−78.5°C) with dimensions of approximately 15 ⨉ 10 ⨉ 3 cm³, was placed on top of the stack. Actual tissue temperatures were measured at 5 locations in the different layers using thermocouples (Omega Inc., Stamford, CO) embedded within carbon-fiber needles. The thermocouples were connected to a data acquisition device (Omega Inc., Stamford, CO) which was placed outside the MR room and sampled at 1 frame/second. To reduce the effect of noise due to MRI gradient-switching on the temperature measurements, three ferrites were attached to each cable. Additionally, a digital second-order Butterworth low-pass filter, with a cut-off frequency of 0.00628 radians/sample, was applied to the acquired signal to further remove high-frequency noise.

Fig. 2.

The apparatus used in the ex vivo tissue experiment. A dry-ice pack was placed on top of a stack of ex vivo swine muscle slices in a Styrofoam container. Five thermocouples embedded in carbon-fiber tubes were inserted into the tissues at different layers. The thermocouples were connected to a data acquisition device (DAQ) located outside the MRI room via cables with attached ferrites. The dry-ice pack gradually cooled down the tissue sample from the top down, forming a temperature gradient within the tissue.

The sample was continuously scanned in the MRI using the scanner’s Body Matrix and Spine Matrix coils using (1) the STIR-UTE-PETRA sequence, and (2) a conventional T2-weighted Turbo Spin Echo (TSE) sequence (TR/TE = 2430/104ms, flip angle = 120°, echo train length = 27, matrix = 192⨉134, pixel size = 0.9375² mm², slice thickness = 3mm, slice gap = 0.6mm, FOV = 180⨉157 mm²).

An intensity bias field was estimated from an image acquired prior to freezing the tissues using the same technique as described in the previous section. Afterwards, the intensity biases in subsequent images were corrected using this estimated bias field. Additionally, the image intensities were normalized to the mean background intensity. The mean pixel intensity in a region of interest (ROI), defined around the tip of each thermocouple, was calculated using the 3D Slicer Label Statistics module, and correlated with the reference temperature obtained from the corresponding thermocouple at the time of image acquisition.

For comparison, the mean pixel intensities in each ROI was also obtained from the conventional T₂-weighted images. The signal intensities were plotted against the reference temperatures.

Clinical Evaluation

To test the feasibility of the proposed temperature-sensitive imaging, STIR-UTE images, along with conventional T₂-weighted TSE images, were acquired during clinical MRI-guided focal prostate cryoablation. Assuming that isotherms surrounding the cryoprobe shaft should appear as ellipsoids, our hypothesis was that a STIR-UTE image obtained perpendicular to the cryoprobe shaft would exhibit a doughnut-shaped hyperintense area, approximately representing the −40°C and −8°C isotherms as the inner and outer rims. Since it was not practical to obtain reference temperatures in vivo during this feasibility study, we only assessed the volumes of the frozen tissue within the inner rim (approximately below −40°C) and the outer rim (approximately below −8°C isotherms), defined as V_inner and V_outer, respectively, on the STIR-UTE images. We hypothesized that V_inner and V_outer would be significantly smaller than the iceball volume, conventionally measured as the void volume on conventional T₂-weighted images.

Data collection

The study was approved by the Institutional Review Board (IRB) and is HIPAA compliant. MR images were acquired in 12 patients (63-80 years old) who underwent an MRI-guided focal prostate cryoablation procedure in the 3T MRI scanner. Cryoablation was performed with an MRI-compatible cryoablation system (SeedNet MRI, Galil Medical Ltd., Yokneam, Israel), utilizing two to five 17-gauge cryoablation probes (Ice Seed, Ice Rod, and/or IceSphere, Galil Medical Ltd., Yokneam, Israel) per patient, depending on tumor volume and geometry. The procedure involved two cycles of 15-minute freezing, interleaved by a 10-minute rapid thawing cycle. All patients were treated under monitored anesthesia care. The iceball was continuously monitored every 3 minutes using a conventional T2-weighted TSE sequence with the same parameters as the frozen tissue experiment as part of clinical routine. Additionally, a STIR-UTE PETRA volumetric acquisition with the same parameters used in the previous experiments, except for voxel size, FOV, and TA (voxel size = 3.4³ mm³, FOV = 330³ mm³, TA=90 sec/vol) was performed near the end of each freezing cycle, if possible, immediately following the conventional T2-weighted image acquisition. All images were exported as DICOM files, and de-identified on a research workstation before further analysis.

Volume measurement of the frozen tissue inside TSE-derived iceball, and STIR-UTE-derived inner and outer rims

On each STIR-UTE image, the outer and inner rims, which corresponded to the estimated −8 °C and −40 °C isotherms, respectively, were manually contoured using the 3D Slicer Editor tool. For comparison, the signal void on each TSE image was also manually contoured.

The volumes of the areas segmented on the STIR-UTE and TSE images were then calculated using the 3D Slicer Label Statistics module and denoted as V_outer(t_UTE), V_inner(t_UTE), and V_Iceball(t_i), where t_UTE and t_i (i = 1, …, N_TSE) are the time points at which STIR-UTE and TSE imaging was performed within each freeze-thaw cycle. Since the STIR-UTE and TSE images were acquired at different time points, the volume of the iceball at the time of STR-UTE image acquisition (V_Iceball(t_UTE)) was estimated by fitting a second-order polynomial function to V_Iceball(t_i). V_outer(t_UTE), V_inner(t_UTE), and V_Iceball(t_UTE) were then compared using Wilcoxon’s rank sum test. A value of p < 0.05 was considered statistically significant.

Results

Pre-Clinical Evaluation

The reference T₁ relaxation times of the T₁ mapping phantoms, quantified with the STIR-TSE images, were 129.5, 110.0, 93.0, 69.7, 252.2, 232.3, 135.0, 141.1, and 168.5 ms. Figure 3 shows STIR-TSE and STIR-UTE images of the T1 mapping phantoms, along with the plot of the mean signal intensity of each vial against T₁ relaxation time. The image confirms that the signal intensity drops as the T₁ relaxation time approaches T_{1, −40°C} = 271ms. As a reference, an STIR-TSE image that shows the closest contrast is also shown along with the signal intensity plot.

Fig. 3.

STIR-TSE (TI=198ms) (left) and STIR-UTE (TI=80ms) (center) images of T1-mapping phantoms consisting of nine vials containing gels with different concentrations. The mean signal intensity in each vial is plotted against the T1 relaxation time for both images (right). The T1 relaxation times, which are shown below each vial on the images, were calculated by fitting the T1 relaxation model to the STIR-TSE images obtained at multiple TIs. The plot indicates that the signal intensity would become zero on both STIR-TSE and STIR-UTE near the expected T₁ (T_{1, −40°C}= 271 ms), but that TSE cannot be used to image frozen tissue, since its TE is too long. The null point for TSE with a TI of 198 ms is expected to be T1 = 198 / ln (2) = 286 ms.

For validation in the frozen tissue sample, a total of 38 STIR-UTE images and 38 TSE images were acquired over a period of 160 minutes. Exemplary images are shown (Figure 4). It is clear from figure 4 that whereas the TSE images only show completely unfrozen tissue (TEMP>~0°C). the STIR-UTE images show only a portion of the frozen tissue, i.e. they provide a temperature selective “ice-image”.

Fig 4.

STIR-UTE (left) and TSE (right) images of the tissue sample (swine muscle) were acquired approximately 40 minutes after the dry ice was placed on top of the sample (see Figure 2 for the configuration). The dry ice created a layer of frozen tissue above the unfrozen layer. The frozen layer is shown as a signal void on the TSE image, while the STIR-UTE image shows a band of hyperintensity between lower temperature and upper temperature boundaries (approximately −40°C<Temp<−8°C) within the frozen tissue. The signal was gradually suppressed near the dry ice corresponding to the decay slope below the lower temperature boundary in Figure 5. The signal from the unfrozen layer is also suppressed.

The mean signal intensity was obtained at 38 time points in five ROIs (total 190 data points / sequence). The relationships between mean signal intensity and temperature for TSE (Figure 5A) and STIR-UTE (Figure 5B) are shown. By comparing Figure 5A and 5B, it is possible to see that the lowest temperature for detection of signal on TSE is ~0°C, whereas the upper threshold for STIR-UTE is ~−8°C, which suggests that there will be a difference between V_outer and V_Iceball. In addition, it would be possible to visually delineate outside the lower and upper boundary temperatures (−40°C and −8°C respectively) on the STIR-UTE images because of the ~−3.2%/°C change in intensity that occurs at these boundaries.

Fig. 5.

Image intensities in frozen and un-frozen muscle tissue on TSE (A) and STIR UTE (B), as a function of temperature. Images are plotted against temperature, measured with thermocouples embedded into the tissue. Arrows indicate lower (−40°C) and upper (−8°C) temperatures where T₁ was expected to be 271ms, which were selected notch temperatures. Note that image intensities below and above these temperatures were attenuated.

Clinical Evaluation

A total of 21 STIR-UTE images were successfully acquired in 12 clinical cases. In three cases, STIR-UTE image acquisition was skipped during one freezing cycle due to interruption of the freezing process. Three to seven conventional TSE images per freezing cycle were acquired during the same procedures. Representative STIR-UTE images and corresponding TSE images, along with 3D renderings of the STR-UTE determined volumes are shown in Figures 6 and 7. The images successfully demonstrated that the method yielded temperature-sensitive contrast where the volume approximately between −40 to −8°C (the periphery) within the iceball appeared as a hyperintense area. Figure 7 shows an area of hyperintensity surrounding the urethra, which was warmed by a urethral warmer catheter to protect the urethral wall throughout the freezing cycle.

Fig. 6.

(Left) Representative intraprocedural multi-slice T2-weighted TSE image acquired after 10 minutes of freezing (upper row) and the corresponding 3D STIR-UTE image (lower row) from Case 1. Original images were acquired in the axial plane (left column) and reformatted along the coronal (middle column) and sagittal planes (right column). Two cryoprobes were placed in parallel to the coronal plane, as indicated by the dotted lines. The STIR-UTE images show a doughnut-shaped hyperintense area within the iceball, approximately representing the −8 °C (outer edge) and −40 °C (inner edge) isotherms. (Right) Three-dimensional renderings of V_outer (Light Blue and Purple) and V_inner (Purple alone) reconstructed from the STIR-UTE images are also shown, along with models of the prostate gland, the cryoprobes, and the urethral warming catheter. The coronal STIR-UTE image and the 3D renderings clearly demonstrate the “synergistic effect”, obtained when iceballs, created by the two cryoprobes, fuse together, thus providing a larger low-temperature region.

Fig. 7.

(Left) Multi-slice T2-weighted (upper row) and 3D STIR-UTE (lower row) images acquired after 12 minutes of freezing from Case 11. (Right) Three-dimensional renderings of V_outer and V_inner reconstructed from the STIR-UTE image are shown, along with models of the prostate gland, cryoprobes, and urethral warming catheter (same notations as in Fig. 6). A large portion of the prostate gland involving the urethra was ablated with five cryoablation probes, resulting in a large signal void on the T2-weighted images. The STIR-UTE images show a hypointense region around the urethra (white arrows), indicating a higher tissue temperature, which is because of the catheter with circulating warm water that was placed within the urethra to protect it from cryoinjury.

The volumes of the iceballs on the TSE image, and the volumes within the −8 °C and −40 °C isotherms on the STIR-UTE images are tabulated in Table 1. V_outer was statistically smaller than V_Iceball (p<1.0⨉10⁻⁶), while V_inner was statistically smaller than V_outer (p<1.0⨉10⁻⁶). The ratios of V_outer to V_Iceball, and V_inner to V_Iceball were 0.92 ± 0.08 and 0.29 ± 0.07, respectively.

Table 1.

Types of cryoprobes and resultant volumes of the iceball on the conventional TSE images and the volumes approximately betwee −8 °C and −40 °C isotherms on STIR-UTE images denoted as V_Iceball, V_outer, and V_inner. The ratios to V_Iceball are also shown in the right two columns.

Case	Probes Utilized	Freezing Cycle	VIceball (cc)	Vouter (cc)	Vinner (cc)	Vouter / VIceball	Vinner / VIceball
1	Ice Seed × 2	1	16.37	15.76	4.14	0.962	0.253
		2	18.25	17.79	4.39	0.975	0.240
2	Ice Seed × 2	1	20.83	14.22	4.43	0.683	0.213
		2	18.86	17.87	5.08	0.947	0.269
3	Ice Seed × 3	1	26.66	19.86	3.82	0.745	0.143
		2	30.92	30.30	7.03	0.980	0.227
4	Ice Seed × 2	2	26.55	23.52	7.07	0.886	0.266
5	Ice Seed × 3	1	28.25	27.78	7.31	0.983	0.259
		2	31.97	31.72	12.27	0.992	0.384
6	Ice Seed × 3	1	46.57	46.02	19.50	0.988	0.419
		2	60.59	56.01	22.26	0.925	0.367
7	Ice Seed × 2	1	31.81	30.02	9.95	0.944	0.313
		2	35.32	34.00	11.05	0.963	0.313
8	Ice Seed × 2	1	18.09	16.90	5.73	0.934	0.317
		2	22.75	19.09	5.28	0.839	0.232
9	Ice Seed × 2	1	19.39	18.89	6.50	0.974	0.335
10	Ice Rod × 3	1	31.55	29.81	10.52	0.945	0.333
		2	32.14	28.76	7.51	0.895	0.234
11	Ice spheres × 4 & Ice Rod × 1	1	61.43	53.90	24.05	0.877	0.391
12	Ice Seed × 2	1	18.33	17.75	5.00	0.968	0.273
		2	20.97	17.06	5.61	0.813	0.267

Additionally, in one patient (Case 2), a postoperative 3D Gd-DTPA contrast-enhanced fast gradient-echo scan (TR/TE = 4.2/1.38 ms, flip angle = 12°, voxel size = 0.5469⨉0.5469⨉2.5 mm³, FOV = 280⨉280⨉90 mm³, matrix = 288⨉140⨉36) was obtained in a 3 Tesla MRI scanner (Discovery MR750w, GE Healthcare, Milwaukee, WI) 24 hours after the cryoablation. Due to different patient positioning in the scanner during the follow-up scan, the image was manually aligned to the intraprocedural MRI for comparison (Fig. 8). The unenhanced (e.g. non-perfused) volume was manually segmented, and considered to be the true cryo-ablated zone. The volume of the ablated zone on the postprocedural CE image was 6.20 cc, whereas the V_outer and V_inner on the STIR-UTE images were 17.87 cc and 5.08 cc, respectively. This suggests a close correlation between V_inner and the true necrotic region.

Fig. 8.

For case 2, Intraprocedural TSE (left), and STIR-UTE (middle) images near the end of the freezing cycle and a postprocedural contrast-enhanced (CE) MRI (right) at corresponding slice locations. The postprocedural CE MRI was acquired in a separate session, and was aligned to the intraprocedural images manually. The boundaries near the −8 °C isotherm (white broken line) and the −40 °C isotherm (blue solid line) contoured on the STIR-UTE image were superimposed on all the images. Additionally, the cryo-ablated zone, estimated from the unenhanced area on the postprocedural CE image (dotted orange line) was also superimposed. The volume of the ablated zone was 6.20 cc, while the maximum volumes within the boundary near the −40 °C isotherm on the STIR-UTE was 5.08 cc.

Discussion

The study demonstrated that STIR-UTE images provided positive contrast in the frozen tissue approximately betwee −8 °C and −40 °C isotherms. While the estimated lethal temperatures may vary between −20 to - 40°C ²¹, this method permits adjusting the boundary temperatures in which hyperintense contrast are obtained by changing the selected TI of the STIR-UTE sequence. The method could therefore potentially provide an alternative approach to MR-based cryoablation monitoring. As compared to other approaches ^22-25, which require a quantitative measurement of R₂* or intensity changes, which must then be converted to temperature based on prior calibrations, this method provides direct visualization of frozen tissue likely above the lethal temperature and therefore offers a simple way to confirm ablation margins without any post processing.

The clinical evaluation demonstrated that the volumes within the outer and inner rims (near the −8 °C and −40 °C isotherms) of STIR-UTE were consistently smaller than those within the signal void of the TSE image, providing additional evidence that the method worked reasonably well in vivo. While we could not find human data showing the volume within the −40 °C cryoablation isotherm, a study on liver cryoablation in an animal model reported that the mean diameters of the −38 °C and 0 °C isotherms were 37 mm and 59 mm, respectively ³⁴. The ratio of the volume within the −38 °C isotherm to the iceball in that study was 0.24, which is close to the ratio of the volume within the inner rim (near the −40 °C isotherm) to the iceball volume (0.29 ± 0.07) we observed. Those findings further support the argument that the proposed method can be used to delineate the volume which is cooled to temperatures below the lethal temperature.

The preliminary comparison between the intraprocedural MRI and the postprocedural CE MRI for Case 2 (Figure 8) demonstrated that the cryo-ablated zone was significantly smaller than the signal void on intraprocedural TSE image and the outer rim (near the −8 °C isotherm) on the STIR-UTE image, and was close to that seen on the inner rim (near the −40 °C isotherm). In fact, the volume within the inner rim (5.08 cc in the second freezing cycle) predicted quite well the volume of the ablated zone (6.20 cc). Unfortunately, validation against the ablation volume was limited to one case in this study, because postprocedural MRI within 24 hours after cryoablation was not available for most cases in this patient cohort.

The study revealed a few potential technical challenges. While Figure 5 clearly demonstrates that STIR-UTE images present high signal intensity within the predicted temperature range in the frozen tissue (Figure 5B), the decay slopes outside the range may obscure the edge of the high intensity area making it difficult to visually delineate the exact isotherms on the image. In our case, the temperature range between the inner and outer boundaries appears to be broader than −40 °C < Temp < 8 °C. Therefore the decay slope could be taken into consideration in the determination of threshold T₁ in future studies. In practice, however, the delineation error due to the decay slope would be less than the pixel size in the clinical setting as demonstrated in the clinical study; given the initial decay slope (~−3.2%/°C in Figure 5) and the relatively steep temperature gradient (~6 °C/mm near the −40 °C isotherm ³⁴, the signal intensity would drop by a third within 2 mm from the −40 °C isotherm.

Another potential challenge is the dependence on the T₁ relaxation time of the frozen tissue. While our goal was to determine the human prostate volume cooled to below the lethal temperature, we used a target T₁ relaxation time obtained from rat muscle ²⁸. While there is no T₁ data for frozen muscle and the prostate were available, we expect that the difference in target T1 relaxation time between the rat muscle and the human prostate is ignorable in practice, given the small dependency of T1 relaxation time on tissue type ³⁵. This assumption was used in a previous study on MR thermometry in frozen tissue ³⁶.

There are also several limitations to this study. The specific PETRA sequence implementation had a spatial resolution of 3.4 mm³, which is relatively large compared to the known cryo-ablation margin (5-10mm). The PETRA k-space sampling scheme also caused strong intensity variations and spatial artifacts, which were not corrected during reconstruction, although it is now possible to correct these with random sampling and SPIRIT reconstruction ³⁷, which might also permit increasing the spatial resolution for the same acquisition time. Those issues can also be addressed by using different types of UTE sequences, such as the 3D UTE AWSOS sequence ³⁸.

The T₁-mapping phantoms used in this study was not suitable for validating the characteristics of the sequence for the proposed application at 3T , because the phantoms did not cover T₁ beyond the threshold T₁ of 271 ms. Another limitation is the use of 2D multislice T2-weighted sequence for the comparison in the clinical evaluation. A T1-weighted 3D gradient echo sequence may be more suitable for accurate volume comparison because of its ability to acquire images with relatively high spatial resolution and short scan time. We selected the T2-weighted TSE sequence to maintain our clinical protocol, which prioritizes soft-tissue contrast outside the iceball over acquisition time.

Finally, this study assumed that the −40°C isotherm at the end of the freezing cycle was the edge of the ablated zone. However, an in vitro study has shown that cancer type, the number of freezing-thawing cycles, the duration of freezing, and the thawing method (active or passive) affect tissue viability post cryo-ablation, in addition to the freezing temperature²¹. Therefore, continuous acquisition of STIR-UTE throughout the freezing process, combined with an assessment of the ‘thermal history’ of the frozen tissue, might further improve the accuracy of the cryo-ablation monitoring. As a result, the −40°C isotherm probably represents a lower bound on the necrosis volume.

Conclusion

In conclusion, we demonstrated a novel approach to predicting the lethal ablation zone using a STIR-UTE MRI. The study showed that the STIR-UTE image successfully delineated the area approximately betwee −40 °C and −8 °C isotherms in the frozen tissue ex vivo and in vivo studies demonstrating its potential as a tool to monitor the lethal ablation area in real-time during MR-guided cryoablation in the prostate and potentially other organs.