Magnetic resonance imaging/transrectal ultrasonography fusion guided seed placement in a phantom: Accuracy between 2-seed versus 1-seed strategies

Abstract

Purpose:

To investigate whether the 2-seed placement per Magnetic Resonance Imaging (MRI) suspicious lesion yields a higher seed placement accuracy than a 1-seed strategy on a phantom.

Methods:

Eight olives embedded in gelatin, each simulating a prostate, underwent MRI. Three virtual spherical lesions (3, 5, and 8 mm diameters) were marked in each olive on the MRI images and co-registered to the MRI/ Transrectal Ultrasonography (TRUS) fusion biopsy system. Two radiologists placed 0.5 mm fiducials, targeting the center of each virtual lesion under fusion image guidance. Half of the 8 olives in each phantom were assigned either to the 1-seed or 2-seeds per lesion strategy. Post-procedure Computed Tomography (CT) images identified each seed and were fused with MR to localize each virtual lesion and collected the seed placement error - distance between the virtual target and the corresponding seed (using the closer seed for the 2-seed strategy). Seed placement success is defined as fiducial placement within a lesion boundary.

Results:

Each operator repeated the procedure on three different phantoms, and data from 209 seeds placed for 137 lesions were analyzed, with an overall error of 3.03 ± 1.52 mm. The operator skill, operator phantom procedural experience, lesion size, and number of seeds, were independently associated with the seed placement error. Seed placement success rate was higher for the 2-seed group compared to 1-seed, although the difference was not statistically significant.

Conclusions:

Placing 2 seeds per MRI lesion yielded a significantly lower error compared to 1-seed strategy, although seed placement success rate was not significantly different.

1. Introduction

Prostate Cancer (PCa) is the most common cancer and second leading cause of cancer-related deaths in men in the United States [1]. Accurate detection of clinically significant PCa at an early stage allows for disease treatment with curable intent in the hope of reducing mortality and improving quality of life [2,3]. Among the many studied techniques to improve the detection rate of clinically significant PCa, Magnetic Resonance Imaging/Transrectal Ultrasonography (MRI/ TRUS) fusion guided biopsy has shown promise to increase cancer detection rate compared to nontargeted systematic biopsy alone [4,5]. In this method, suspicious lesions identified by MRI are spatially mapped to real-time transrectal ultrasound images at the time of biopsy to facilitate targeted tissue sampling of the highest risk areas within the prostate [6]. In addition, accurate fusion MRI/TRUS guidance can improve risk stratification of prostate lesions by assessing tumor characteristics such as core involvement percentage and core tumor length [7].

The optimal number of biopsy core samples for each MRI-detected lesion remains unclear. While the currently accepted strategy for MRI/ TRUS guided biopsy includes obtaining 2 tissue cores for each suspicious lesion on MRI [8,9], there is no convincing evidence to suggest superiority of two cores versus a single sample. A recent retrospective clinical study found that multiple core samples do not significantly improve the cancer detection rate or change the Gleason scores, compared to obtaining a single core biopsy per MRI suspicious lesion [10], however, they did not fully analyze the influence of potential confounders for prostate biopsy accuracy such as lesion size, lesion location, and operator skills [11–14].

Due to cost, ethical issues, and safety concerns, the optimal targeted core sample number has not been fully investigated in vivo. Recently, our group created a configurable gelatin-based phantom and demonstrated its application in simulating interventional MRI/TRUS procedures with the goal of testing relevant clinical hypotheses that cannot be studied on human subjects [15]. In this study, we performed MRI/ TRUS fusion guided metallic seed placement in our phantom as a surrogate for targeted prostate biopsy. We quantitate precision of metallic seed placement and investigate whether the placement of 2 seeds per lesion yields a higher seed placement accuracy than a single seed strategy.

2. Materials and methods

2.1. Study overview

Gelatin phantoms with embedded olives to simulate the prostate were constructed and imaged with MRI. After the olives were manually contoured on MR images, virtual lesions were created and then co-registered with ultrasound images. Metallic seed placement was performed, targeting the center of each virtual lesion under the guidance of an MRI/TRUS fusion system. Due to metallic seed-MRI incompatibility, a post-procedure CT was performed to identify the seeds’ location. Initial MRI images were rigidly co-registered to CT images in order to measure the distance between the virtual targets and metallic seeds, the seed placement error. (Fig. 1) IRB approval was not required for this phantom study.

Fig. 1.

Flowchart of the experimental workflow to investigate seed placement error. After the seeds were placed under ultrasound guidance targeting the virtual lesions on MRI, a post-procedure CT with the metallic fiducials was co-registered to MRI to calculate seed placement error.

2.2. Phantom construction and MRI scan

The phantom mold is composed of a cover lid, cylinder body, and base with eight solid poles. Eight large olives (longitudinal axis ranging 2.0–2.5 cm) were placed on the poles, each simulating a prostate. The phantom matrix is made of plain tap water, gelatin, agar, graphite, and 1-propanol based on the recipe established in the previous study [15]. Three MRI markers (MR-SPOT Packets™ 184, Beekley Medical, Bristol, CT, USA) at different orientations were attached to the sidewall of the phantom mold for co-registration of MRI and CT images. Following construction, each phantom was scanned using a 3 T MRI scanner (Magnetom Prisma, Siemens Healthcare Diagnostics, Tarrytown, NY, USA), obtaining the T2-weighted sequence with the following parameters: TR 3200 ms, TE 563 ms, voxel size 0.8 × 0.8 ×0.8 mm, FOV 200 mm, SNR 1.00, flip angle 5 deg.

2.3. Contouring and re-slicing the MRI images

MRI data were loaded and processed using the pre-installed MIM software (Version 6.8.2, MIM Software Inc., Beachwood, OH, USA) on the BK ultrasound machine (BK 3000, Analogic Corp., Peabody, MA, USA). For each olive, the “2D Brush” function was used to contour the olive on every slice, followed by interpolation of the contours to create a 3D annotation of the olive. Subsequently, the “3D Brush” function was used to create three non-overlapping virtual spherical lesions with diameters of 3, 5, and 8 mm (reflecting typical small lesions identifiable by MRI) within the olive. (Fig. 2A)

Fig. 2.

Olive segmentation and lesion markup on MRI images. The interpolated contours of the olive (pink) and virtual lesions (blue, red, and yellow circles) on transverse plane (A). In each phantom, four olives (#1, #3, #5, #7) were assigned for 1-seed placement per lesion and the other four olives were assigned for 2-seed placement per lesion. Each operator started with one olive for 1-seed placement per lesion, followed by another olive using 2-seed strategy per lesion. The ReSlicer indicator on top of the screen was reoriented parallel to the longitudinal axis of the olive (B). The grid was adjusted to ensure that the olive was centered and the bottom contour of the olive was tangential to the first row of the hashes (C).

In order to adjust the orientation of the MR images with the initial alignment of the probe during the procedure, the images were resliced at the appropriate angle by reorienting the ReSlicer indicator parallel to the longitudinal axis of the olive (Fig. 2B) and adjusting the grid in a way that the olive was centered and its bottom contour was tangential to the first row of the hashes (Fig. 2C). After reslicing, the MR images were saved for the designated olive to be used during seed placement.

2.4. MRI/TRUS fusion guided seeds placement protocol

Procedure setup is depicted in Fig. 3A. Metallic seed placement was performed using the BK ultrasound machine with an endocavitary transducer (E14C4T, Analogic Corp., Peabody, MA) and an attached needle guide (REF UA1329-S, Analogic Corp., Peabody, MA). The ultrasound scanning presets were constant throughout the study: depth at 4 cm, B-mode frequency at 12 MHz, B-mode gain at 50 %, and mechanical index at 1.24. To track the spatial location of the probe during the procedure, an electromagnetic (EM) field generator was placed within two feet of the EM sensor clamped to the handle of the ultrasound transducer.

Fig. 3.

MRI/TRUS fusion guided seed placement procedural setup. Seed placement apparatus includes (1) BK MRI/TRUS fusion system, (2) magnetic generator, (3) phantom, and (4) the transrectal probe with the needle guide (A). The center of the virtual lesion (cyan circle) is targeted for seed placement according to the needle guideline (dotted yellow line) following optimal alignment of the olive contour between MR (blue oval line) and ultrasound US images (B). The composite 3D volume rendering shows the ultrasound detected needle (green bar) tip reaching the center of the targeted virtual lesion overlay (cyan sphere) and avoiding other virtual lesions (red and yellow), with the dimmed oval olive overlay as the background (C).

Metallic fiducial markers were prepared by cutting a 24-gauge steel wire into 0.5 mm segments (OOK Galvanized Steel Wire, Home Depot, Watertown, MA). A previously used 18 gauge ×20 cm Visicoil needle (RadioMed Corporation, Bartlett, TN), comprised of a needle sheath and a stylet, was used for seeds delivery.

Two radiologists with more than 5 years’ experience in prostate fiducial placement performed seed placement on three phantoms on three different days. Neither operator had experience with seed placement in phantoms prior to this study; they were instructed on the operational workflow in the first experiment. Each operator started with one olive for 1-seed placement per lesion, followed by another olive using 2-seed strategy per lesion; this procedure was repeated until four olives for 1-seed placement and the other four for 2-seed placement were completed in each phantom. (Fig. 2A)

Seed placement was performed by targeting the center of the virtual lesions within one olive. After loading the resliced fusion data, the operator obtained sweep ultrasound images through the olive and adjusted the contour to ensure the optimal alignment between ultrasound and MRI images. The best matched MRI/TRUS image sets were locked such that seed placement could be guided by the fused images (Fig. 3B). Once the lesion center was visually confirmed on the MRI/US fusion images, the operator stabilized the probe at the optimal position and advanced the needle sheath via the needle guide until the needle tip reached the lesion center. The metallic fiducial marker was placed into the sheath from the top using a small tweezers and pushed into the target using the stylet (Fig. 3C). For the 2-seed placement, the operators were asked to completely withdraw the needle and restart the workflow for the second seed placement.

2.5. MRI/CT co-registration and seeds placement errors calculation

CT scan (Discovery CT750HD, GE HealthCare, Waukesha, USA) of the phantom was performed with scan parameters of 120 kV, 500 mAs, display field of view of 21.5 cm, helical thickness of 1.25 mm, and reconstructed with 1-mm-thickness slices.

The phantom MRI and CT images were then rigidly co-registered on the MIM platform. (Fig. 4A) The virtual lesion center was identified using the “Localize to Contour Centroid” function. (Fig. 4B) By manually adjusting the window width of CT images, each metal seed with artifact was identified and, by raising window width, artifact was minimized to precisely localize the seed’s center. (Fig. 4C)

Fig. 4.

Seed placement error calculation on MRI/CT co-registered images in 1-seed strategy. Axial, sagittal, and coronal (left to right) phantom MRI (top row) and corresponding CT (middle row) images and virtual lesion (red circle) were co-registered (bottom row) on the MIM platform (A). The spatial coordinates of a lesion (yellow circle) center were identified using the “Localize to Contour Centroid” function at (0.63, 30.75, 16.82) when mapped to CT images (B). The threshold of CT value was adjusted around the metal seed to identify the center of the seed (1.87, 31.54, 18.49). The distance (placement error) between the seed center to lesion

Based on coordinates, the seed placement error was computed as the distance of the seed center to the lesion center in the 1-seed group; in the 2-seed group, the seed closer to the lesion center was chosen to represent the seed placement error for each lesion (Fig. 5). The distances between the centers of the two seeds for each lesion was also calculated in the 2-seed group.

Fig. 5.

Seed placement error calculation in 2-seed strategy. In the olive (grey oval), two seeds (black stars) were placed targeting the center (black solid dot) of virtual lesion (red circle). After obtaining the coordinates of two seeds (S1 and S2) and lesion centers, the distances (d1 and d2) of each seed to the lesion center were calculated respectively. As S1 was closer to lesion center, d1 was chosen to represent the placement error for this lesion.

2.6. Statistical analysis

Statistical analyses were performed using SPSS Statistics 25.0 (SPSS Inc., Chicago, USA). Quantitative variables were expressed as mean ± standard deviation (SD), and qualitative variables were expressed as frequencies and percentages. Multiple linear regression was performed to assess the factors affecting seed placement error (distance from lesion center to seed) and precision (distance between two seeds within one lesion). Model independent variables were selected a priori based on previous reports and included operators, phantoms sequences, lesion diameters, and seed number strategies. For each covariate, results were reported as β coefficients with 95 % confidence intervals (CIs). The statistical adjustment for multiple comparisons was not performed as the variables assessed were pre-defined, small in number, and each covariate had a scientifically plausible relationship with the outcomes. Chi-Square analysis was used to compare the seed placement success rates between 1-seed and 2-seed groups. All tests were 2-sided, with a significance level of 0.05.

3. Results

3.1. Seeds placement errors

Each operator placed seeds in 3 phantoms, and a total of 144 virtual lesions in 6 phantoms were targeted for seeds placement. For the 72 lesions in the 1-seed group, 7 seeds (4 by operator 1, 3 by operator 2) were missing on CT; while in the 2-seed group, all 144 seeds were successfully placed into the 72 lesions by both operators. Finally, data from 209 seeds in 137 lesions were analyzed.

Seed placement error (Fig. 6) ranged from 0.09 to 11.59 mm (mean 3.03 ± 1.52 mm). The two largest errors (10.26 and 11.59 mm) occurred in the first phantom of operator 2 using 1-seed strategy, when targeting lesions of 8 mm and 5 mm, respectively. Seed placement errors stratified by operators, phantoms, and lesion diameters were presented in Table 1.

Fig. 6.

Distribution of seed placement error by seed placement strategy. The distribution of errors in both groups was positively skewed, with a mean of 3.10 ± 1.83 mm and a median of 2.74 mm (25th and 75th percentiles: 2.33, 3.40 mm) in 1-seed strategy (blue bars), and a mean of 2.54 ± 1.17 mm with a median of 2.46 mm (25th and 75th percentiles : 1.83, 3.15 mm) in 2-seed strategy (red bars).

Table 1.

Seeds placement error stratified by covariates.

Error Sources	Virtual Lesion Numbers (n=137)§	Seeds Placement Error (mm)*
		1-Seed Strategy (n=65)	2-Seed Strategy (n=72)	p value
Operator
Operator 1	68	2.54 ± 0.99	2.28 ± 1.18	0.418
Operator 2	69	3.64 ± 2.25	2.79 ± 1.13	0.312
Phantom sequence
#1	44	4.31 ± 2.67	2.95 ± 1.55	0.293
#2	45	2.39 ± 1.07	2.39 ± 0.89	0.134
#3	48	2.70 ± 0.68	2.26 ± 0.88	0.318
Lesion Diameter
3 mm	46	2.35 ± 0.90	2.14 ± 0.83	0.592
5 mm	46	3.25 ± 2.23	2.83 ± 1.52	0.812
8 mm	45	3.72 ± 1.89	2.63 ± 1.01	0.127

^*Mean ± standard deviation.

^§Virtual Lesion Numbers: virtual lesions with identified seeds placement on CT.

3.2. Incremental success rates of seed placement within different distances to lesion centers

Success rate of seed placement was defined as the percentage of seeds successfully placed within a specific threshold distance from the center of the targeted lesion. For the 2-seed strategy, the placement was considered successful as long as one of the two seeds was placed within the specific distance from the lesion center. Expectedly, success rate increased by increasing the threshold distance from the lesion center in both 1-seed and 2-seed groups without significant difference between the groups (p > 0.05). The greatest changes in success rate, 18.5 %–84.6 % for 1-seed strategy and 30.6 %–88.9 % for 2-seed strategy, were observed when the distance increased from 2 mm to 4 mm. The success rates increased in a sigmoidal manner and reached 100 % at 7 mm in 2-seed group, and at 12 mm in 1-seed group. (Fig. 7)

Fig. 7.

The incremental success rates using different threshold distances from lesion center. The success rates increased in a sigmoidal manner with increasing threshold distance in both 1-seed (blue solid line) and 2-seed groups (orange solid line). The greatest change was observed when the distance increased from 2 mm to 4 mm.

3.3. Factors affecting seed placement errors

Multiple linear regression was performed to identify the factors affecting seed placement error. (Table 2) Each variable assessed was found to associate with seed placement error when controlling for other covariates. Seed placement error was significantly higher for the seeds placed by operator 2 and when targeting the larger virtual lesions. On the other hand, the error was smaller in the second and third phantoms compared to the first phantom. Also, the error was significantly lower in the 2-seed strategy, compared to 1-seed placement (β coefficient = −0.588, p= 0.011).

Table 2.

Factors affecting seed placement error.

Factors	β Coefficient (95 % CI)	p Value
Operator
Operator 1*	–	–
Operator 2	0.729 (0.277,1.181)	0.002
Phantom sequence
# 1*	–	–
# 2	−1.136 (−1.697, −0.575)	< 0.001
# 3	−1.093 (−1.644, −0.541)	< 0.001
Lesion diameter
3 mm*	–	–
5 mm	0.784 (0.234, 1.335)	0.006
8 mm	0.912 (0.358, 1.465)	0.001
Seed placement strategy
1-seed*	–	–
2-seed	−0.588 (−1.040, −0.136)	0.011

CI: confidence interval.

^*Reference levels of variables in the multiple regression analysis.

3.4. Inter-seed distance in 2-seed group

The distance between two seeds within the same lesion (range 0.43–8.68 mm) was calculated to analyze the precision of seed placement. Multiple linear regression analysis showed the inter-seed distance to be significantly lower in the second and third phantoms compared to the first phantom (#1: 2.91 ± 1.79 mm, #2: 1.66 ± 0.85 mm, #3: 1.26 ± 0.67 mm) indicating that precision improves with training, regardless of the operator and lesion size. (Table 3)

Table 3.

Factors affecting seed placement precision in 2-seed strategy.

Error Sources	Inter-seed Distance (mm)*	Multivariate Linear Regression Analysis
		β Coefficient (95 % CI)	p Value
Operators
Operator 1§	2.14 ± 1.45	–
Operator 2	1.75 ± 1.30	−0.395 (−0.961, 0.171)	0.168
Phantom sequences
# 1§	2.91 ± 1.79	–
# 2	1.66 ± 0.85	−1.249 (−1.943, −0.556)	0.001
# 3	1.26 ± 0.67	−1.649 (−2.343, −0.956)	< 0.001
Lesion diameters
3 mm§	1.79 ± 0.95	–
5 mm	2.19 ± 1.93	0.397 (−0.296, 1.090)	0.257
8 mm	1.85 ± 1.08	0.062 (−0.631, 0.755)	0.859

^*Mean ± standard deviation.

^§Reference levels of variables in the multiple regression analysis.

4. Discussion

4.1. Measurement of MRI/TRUS fusion guided seeds placement error

Accurate MRI/TRUS fusion guided biopsy is needed for reliable estimation of core tumor length, which is well correlated with Gleason scores [16]. Thus, minimization of phantom seed placement error, as a surrogate marker of biopsy accuracy, is expected to correlate with improved risk stratification of prostate cancer.

Due to the small size of the metal seeds (∼0.5 mm), a total of 7 seeds from the first phantom were missing on the post-procedure CT scans, likely related to technical challenges pertaining to operator inexperience. Subsequent phantoms were without missing seeds and the overall success rate of seed placement (209/216, 96.8 %) was acceptable. In the clinical setting, when multiple cores were obtained from one prostatic lesion, the most representative one was usually used for pathological diagnosis, so we chose the seed closer to the lesion center to represent the seed placement error for each lesion in the 2-seed group. [17] Our overall MRI/TRUS seed placement error (3.03 mm) was comparable to previous phantom data (2–4 mm) reported by other groups [13,18–20]. As shown in Fig. 6, two outliers more than 10 mm in the first phantom considerably increased the average error, but the majority of distance values in subsequent phantoms were reasonably low (2.30–2.70 mm).

A wide range of different error sources have been reported during the MRI/TRUS fusion biopsy, including image registration error, procedural errors, needle deflection error, physician operational error, and others. [21] While our study did not directly measure each of these factors, we directly computed overall accuracy and control for several confounding factors (operator skill, procedural experience, lesion size, etc.). Controlling all the confounding factors in ideal conditions would enable reliable comparison between single or multiple biopsy strategies.

4.2. Influencing factors for the MRI/TRUS fusion guided seeds placement error

Our study showed that the operator skill, procedural experience, lesion size, and seed placement strategy are independently correlated with seed placement error. Therefore, the accuracy of the MRI/TRUS seed placement relies on appropriate management of these influencing factors during the procedure.

Congruent with another report [22], we found a higher accuracy in the 2-seed group compared to the 1-seed group suggesting that additional core biopsies have the potential to improve the probability of tumor sampling. This may provide more accurate quantitative tumor measures in prostate needle biopsy specimens, such as percentage of a biopsy core with cancer. Considering the challenges in controlling multiple sources of error during seed placement, additional core needle samples may be a solution to mitigate sampling error.

As mentioned, phantom sequence was independently associated with seed placement error, which was an indicator of a learning curve of seeds placement for both operators. Both operators in this experiment had no prior operational experience on the current phantom, with the experience gained in the first trial (error of 4.31 and 2.95 mm), their second and third attempts were more successful (error of ∼2.39 mm).

It’s important to note that, the two largest errors (10.26 and 11.59 mm) occurred in the first phantom of operator 2, which maybe one potential reason to explain his overall lower accuracy in the first phantom. Regarding the influence of the operators, even though the two board-certified interventional radiologists had similar clinical training background, intra-operator variations led to a significant difference between their performance on the first phantom, but the gap decreased after further training on the next phantoms.

Although perhaps counterintuitive, seed placement error increased for larger target lesions. This may be explained by increased difficulty in visual identification of the target center for larger targets, introducing more operator-dependent errors during seed placement. For larger lesions, small movements of the probe are more difficult to visually detect and may go unnoticed, while for smaller lesions, the slightest movement can make obvious changes in lesion visualization alerting the operator to correct the seed placement trajectory. While more difficult to accurately target, this introduced error for sampling larger lesions may be mitigated by the large target size.

4.3. Success rates of seeds placement at incremental distance thresholds to target lesion centers

The cancer detection rate per core, or lesion hit rate, is the most commonly used parameter for evaluating the accuracy of a biopsy plan. The success rate of seed placement in this study was related to this parameter. Consistent with the previous phantom study, our results showed the success rate to incrementally increase with the increasing threshold distances to the lesion center in both 1-seed and 2-seed groups, and lesions larger than 10 mm (corresponding to an allowable error of 5 mm) would be successfully biopsied in more than 90 % of cases (Fig. 7). PIRADS 5 lesions on MRI (greater than 15 mm) [23] would be detected in 100 % of our simulated biopsies. The sigmoidal curve showed that the success rate increased with the greatest change when the distance increased from 2mm to 4 mm (corresponding to 4–8 mm lesions), which indicated the substantial decrease in accuracy occurs for small lesions, especially for those smaller than 4 mm.

Although we were not able to detect a statistically significant difference between the success rates of the 1-seed and 2-seed strategies, there was still a trend toward improved successful diagnosis for the 2-seed strategy especially for small lesions (4–8 mm in diameter). However, the value of 2-seed strategy remains controversial. Some previous clinical studies showed that the benefit of a second targeted biopsy core per MRI-lesion was likely minor [10,24]. On the contrary, another phantom study reported that acquiring more cores could provide complementary accuracy, but they calculated the likelihood of the 2nd core’s success from the data obtained from the first core using a probabilistic formula [20], rather than actually performing the second core test.

4.4. Inter-seed distance analysis in 2-seed group

The overall errors during the MRI/TRUS fusion guided procedures generally include systematic and random errors [25]. The analysis of inter-seed distance in the 2-seed group provided the error information mainly from the system itself, which ranged 1.26–2.91 mm in our six phantoms. Identification of the inter-seed distance can gain insight into the system performance for the purpose of quality assurance and system improvement. Our multivariable regression analysis showed the inter-seed distance was not associated with the operator nor lesion size, but decreased in consecutive phantoms while controlling for other covariates, suggesting that the biopsy outputs of the 2-seed strategy can be alternatively improved with operator training.

4.5. Limitations

Although this phantom study provides an opportunity to test specific prostate biopsy strategies, these results may be difficult to generalize to the in vivo setting due to differences between the properties of phantom and human tissues, such as the tissue rigidness, organ shape, organ size, and in vivo pathology such as prostate calcification, all of which may impact biopsy performance. However, the ability to control sources of error sources in phantoms, often not possible under in vivo conditions, provides insight into the performance of a fusion system under idealized conditions. Second, we did not use core needle biopsy to confirm accurate needle trajectory. Instead, we used seed placement in virtual lesions as a surrogate for lesion biopsy. This strategy allows the quantification of needle placement precision, which is not possible with the biopsy strategy. Estimation of inter-seed placement error is limited by the physical size of the seeds which precludes an error smaller than 0.5 mm (seed size). This limitation is not applicable to other error estimation since the physical seeds could occupy the same space as the virtual lesions. Needle placement accuracy alone may not account for all variation in clinical biopsy success. Additional factors, such as needle gauge and biopsy device may play a role, however this could not be assessed by seed placement and was beyond the scope of this study. In subsequent experiments, we will directly assess the efficacy of core biopsy by assessing focal lesion biopsy success rate though sampling of phantom-embedded lesions. Assessment only of spherical lesions of three fixed sizes may not reflect the clinical reality of real prostate lesions which are more diverse in both size and shape. Limited tools in the BK fusion system did not allow the creation of virtual lesions with varied shape. Limited space within each olive precluded assessment of more than three distinct lesion sizes. Additional study is warranted to address the role of lesion size and shape on biopsy precision.

5. Conclusions

In this phantom study, we compared the overall accuracy and success rate of seed placement between the 1-seed and 2-seed strategies. Our results show that the 2-seed scheme per target lesion yielded a significantly lower error compared with the single seed strategy. While this did not result in improved virtual biopsy success rates across all lesion sizes, 2-seed placement may be particularly beneficial for small target lesions where biopsy accuracy suffers from sampling error. These preliminary findings may be helpful to optimize the procedure strategies of MRI/TRUS targeted prostate biopsy.