Simple Frame-based Approach for MR Imaging-guided Abdominal Biopsy: Preliminary Evaluation in a Phantom Model

Abstract

Purpose

Although automation offers several benefits in MRI-guided abdominal biopsies, the conventional surgeon-centric approach continues to be the preferred technique. This study presents an ergonomic biopsy device aimed at streamlining the MRI-guided abdominal biopsy process while preserving operator control.

Methods

A dedicated belt was designed with adjustable straps to accommodate patients of varying sizes. A biopsy frame with 2 manual degrees of freedom was movably affixed to the belt, allowing for adjustable positioning across the patient’s circumference to align with the region of interest. Preliminary testing was performed on an abdominal biopsy phantom with 2 embedded 7-mm tumor mimics in a 3T MRI scanner. Needle guide registration in MRI coordinates was achieved by attaching a water-filled syringe to the guide, which was then navigated to align with each target using a straightforward method involving the acquisition and fusion of parallel images. A clinically relevant half-Fourier acquisition single-shot turbo spin echo sequence was used for both the initial alignment and for intermittent imaging during the stepwise advancement of a metallic clinical needle toward the target.

Results

The device was easy to use, even for a non-experienced operator. The employed methodology enabled straightforward calculation of the 2D path for aligning the syringe with the target. Alignment accuracy was within the submillimeter range. The employed sequence was robust against susceptibility artifacts, enabling precise guidance, with the accuracy of tumor puncture ultimately dependent on the operator.

Conclusion

The proposed biopsy method has the potential to enhance stability and accuracy over traditional techniques, while being more ergonomic and cost-effective than advanced robotic systems, also maintaining human control. Extensive preclinical and clinical evaluation is needed to fully assess effectiveness and safety.

Introduction

Ultrasound (US) guidance is traditionally favored for sampling superficial abdominal wall lesions due to its accessibility and cost-effectiveness, though CT may be required for very small or deeply located tumors.^1,2 Meanwhile, MRI, with its superior soft tissue contrast and lack of ionizing radiation, is increasingly favored for diagnosing abdominal wall masses, offering enhanced diagnostic value.³

US-guided robotic systems have been proposed to improve the accuracy of percutaneous needle insertions and streamline the biopsy process by reducing dependence on operator skill, with various phantom studies demonstrating their effectiveness.^4–8 Similarly, CT-guided biopsies have benefitted from technological innovations. For instance, ANT-C (NDR Medical Technology, Singapore) utilizes artificial intelligence to analyze CT scan images, automating lesion detection, needle path planning and targeting processes.^9,10 Remarkably, robots for CT-guided needle insertions are commercially available in several types, including patient-mounted,¹¹ table-mounted,¹² and floor-mounted¹³ systems. Recent advancements in robotic-assisted percutaneous puncture procedures further include the development of advanced calibration techniques¹⁴ and the exploration of fused imaging methods to enhance targeting accuracy and procedural outcomes.^15,16

The use of MRI to guide abdominal biopsies has notably increased over the past 2 decades, driven by its superior imaging capabilities and patient benefits.^17–20 Building on earlier, less precise methods such as the “finger pointing” technique,^21,22 which involved attaching an MR-visible capsule to the physician’s fingertip,²² later advancements introduced more sophisticated localization methods. These methods typically involve the use of MR-visible reference markers and fiducials, which can be affixed to the patient’s skin, the patient table^18,19 or directly on the biopsy instruments.¹⁸ In the free-hand technique, the radiologist adjusts the needle by hand relying on these reference points for guidance.^23,24 By contrast, frame-based methods provide a stable, fixed reference attached to the patient, typically featuring grid-like reference points that help maintain trajectory consistency, effectively reducing sampling errors.²⁵

The typical process for MRI-guided percutaneous body biopsy involves the following steps: (1) Initial imaging of the ROI to identify the target lesion and assess general anatomy of the area; (2) Attachment of localization tools to the setup or the patient’s skin over the ROI, and/or body-mounted stereotactic frames; (3) Repeated imaging to precisely define the entry point and optimal needle trajectory, taking into account sensitive structures; and (4) Stepwise insertion of the biopsy needle with interstep imaging to monitor progress, typically involving the acquisition of a series of slices parallel to the desired trajectory.^17–20 Notably, in-room monitoring is feasible by installing an MR compatible monitor adjacent to the bore. In a relevant retrospective cohort study,¹⁷ the overall median procedure time for liver and soft tissue biopsies was approximately 105 mins, influenced by delays from needle advancement pauses for confirmatory imaging.

Several clinical investigations on the effectiveness of MRI-guided abdominal and pelvic biopsies are documented in the literature.^18,20,22 One study treated liver lesions in a cohort of 52 patients utilizing an add-on navigation system in a 1.5T scanner, with biopsy instrument manipulation shown on a screen in the procedure room.¹⁸ The system was capable of safely targeting difficult-to-access liver lesions within a mean duration of 51 ± 12 mins, demonstrating an overall diagnostic accuracy of 92%. In another study,²⁰ Kühn et al. report on 50 biopsies of upper abdominal tumors performed in a 3T scanner over a 1-year period. Diagnostic accuracy was reported at 94%, with minor complications in 14% of cases, supporting the effectiveness of performing upper abdominal biopsies in large-bore scanners. Driven by the need to reduce procedural times, Stattaus et al.²² evaluated the feasibility of MR fluoroscopy for near real-time guidance in a short-bore 1.5T scanner, achieving accurate needle placement in 8 patients with liver or soft-tissue tumors, without requiring out-of-bore movement.

Robot-assisted MRI-guided biopsy is an emerging, yet relatively understudied field compared to CT-guided robotic biopsy. In the preclinical landscape, He et al.²⁶ developed a body-mounted robot with soft fluid-driven actuators for semiautomated needle guide positioning for abdominal organs. INNOMOTION (Innomedic Philippsburg, Germany) is the first commercially available MRI-compatible servo-pneumatically driven robot for percutaneous interventions in the MRI environment, with the second generation offering 6 degrees of freedom (DOF) along with an additional passive rotational DOF.^12,27,28 The biopsy instrument serves as the end effector of a robotic arm affixed to the patient table. Evaluation in ex-vivo and in-vivo porcine models, focusing on compatibility and targeting precision in 1.5T scanners,¹² laid the groundwork for CE mark approval. Later assessemet of needle placement accuracy in a plastic brick phantom revealed an average off-target deviation of 2.2 ± 0.7 mm.²⁸ Initial clinical data on abdominal-pelvic biopsies demonstrated diagnostic success with no complications, though with a highly variable intervention time of 36‒68 mins, suggesting further optimization could improve efficiency and reduce procedure duration.²⁸

One could argue that research in MRI-guided abdominal biopsies is progressing slowly, with limited exploration of new systems and tools for ergonomic and time-efficient targeting of abdominal lesions, while nearly all studies on robotic-assisted intervention focus on a single system (INNOMOTION).²⁸ Additionally, despite efforts to improve accuracy and reduce procedural time by automating techniques in closed-bore scanners^26,29,30 or employing open scanners for an uninterrupted process,³¹ traditional methods for MRI-guided biopsies remain the preferred choice. This is further compounded by factors such as the increased complexity and cost of incorporating robotics in the MRI room, challenges of developing compatible robotic systems in terms of size and materials,³¹ lower availability and magnetic strength of open scanners, and physicians’ preference for procedures that maintain human control. Motivated by these factors, we herein present a straightforward approach for MRI-guided abdominal biopsy using a dedicated ergonomic 2-stage biopsy belt, supported by preliminary MRI evaluation on a biopsy static phantom. A metallic clinical needle was used for stepwise target puncture, following needle guide-to-target alignment facilitated by a water-filled syringe under MRI guidance.

Materials and Methods

The study was conducted in a 3T conventional MRI scanner (Magnetom Vida; Siemens Healthineers, Erlangen, Germany), equipped with an Ultraflex small 18-channel body coil to facilitate imaging of an agar-based biopsy phantom throughout all procedural stages.

Design of X-Z stage biopsy belt

A rectangular plastic frame was 3D-printed on a fused deposition modeling (FDM) printer (Raise3D E2 printer; Raise3D, CA, USA) using polylactic acid (PLA) thermoplastic. The frame was designed with 2 manual DOF for adjusting the needle guide position in the Z (anterior-posterior) and X (left-right) directions, in a patient lying in a supine position. Specifically, the frame features a plastic slide that can be moved in the Z direction along dedicated guides on the left and right sides of the frame. This slide in turn incorporates a needle guide that can move along it in the X direction. The motion range is 60 mm in the X direction and 30 mm in the Z direction. It has scale markers engraved at 1 mm intervals to guide manual adjustment of the needle guide position. The entire frame was mounted on adjustable straps, allowing it to be positioned across the abdominal area to control the insertion angle. These straps further allow the user to customize the fit according to the waist size. The various components of the biopsy belt can be seen in the computer-aided design model of Fig. 1a.

Fig. 1.

CAD models of the (a) biopsy belt and (b) abdominal biopsy phantom.CAD, computer-aided design.

Abdominal biopsy phantom

An abdominal tumor-bearing phantom was developed using agar (Merck, Darmstadt, Germany) as the gelling agent. Agar was selected for its common use in MRI phantoms and its ability to produce tissue-like signal.^32–35 Spherical tumor mimics of 7 mm diameter were incorporated to serve as the biopsy targets. Both tumor and normal tissue materials contained a 6% weight/volume (w/v) agar concentration, with an additional 6% w/v silica (Sigma-Aldrich, Missouri, USA) in the tumors to reduce T2 relaxation time and enhance contrast in T2-weighted (T2-W) imaging. The phantom was formed into a semicircular block of 30 cm in length, 12.5 cm in height, and 10 cm in width, by molding in a dedicated plastic mold, to resemble the shape of the abdomen, as shown in Fig. 1b.

Two targets were incorporated within the semicircular block, one positioned at the center, and the other on the left side. They were spaced out 6 cm ($d_1$) and 4.5 cm ($d_2$) from the upper (curved) phantom surface, respectively. The target point was defined in tumor center. The Young’s modulus of this phantom was calculated at 2.4 ± 0.1 GPa (n = 5), providing an intrinsic measure of the material’s stiffness. This value was derived from the speed of sound measured using the pulse-echo method³⁵ and the material’s density.

MRI-guided tumor puncture technique

The experimental setup is illustrated in Fig. 2a and Fig. 2b, the latter highlighting the placement of the coil above the belly phantom, supported by 2 cushions. The coil position was determined in consultation with physician advisors to reflect a clinically relevant configuration, and it was adjusted to remain close to the area of interest, i.e., adjacent to the fixed frame, depending on the target’s location. Two different insertion orientations were tested: one simulating a direct approach into the flat belly and the other representing a lateral (sideways) insertion. The procedure included the following steps:

1. Target Localization: The procedure began by securely fitting the belt to the phantom using the adjustable straps to ensure stability. Multi-plane imaging was then performed to locate the biopsy target.

2. Positioning and Registration: The biopsy frame was approximately positioned over the target area by the operator and registered to the MRI coordinate system. Registration was achieved by positioning the needle guide at a predefined home location, with a water-filled syringe attached to it, serving as a reference on T2-W images.

3. Needle Guide Path Calculation: Two parallel slices, oriented perpendicular to the syringe axis (with the syringe oriented normal to the phantom surface), were acquired at the level of the syringe and phantom; specifically coronal and oblique images for the vertical and lateral insertion approaches, respectively. These images were then fused to calculate the in-plane deviation (in mm) between the needle guide and the tumor simulator center. This planning process is supported by specialized software, which allows the operator to select the target location on preoperative images. The software automatically calculates the 2D (in-plane) displacement required for the needle guide to align with the target, along with the necessary depth to reach the target center.

4. Alignment and Adjustment: Once the trajectory was established, the operator made manual adjustments to the needle guide using the scale markings on the biopsy frame. Additional imaging was performed to verify the accuracy of the adjustment before moving on to the next step.

5. Needle Insertion and Verification: After confirming the alignment, the water-filled syringe was replaced with a coaxial interventional needle (MR Conditional, 16 G/1.6 mm, Length: 144 mm; Innovative Tomography Products, Bochum, Germany). Equipped with graduated depth markings, the needle enabled the operator to gauge insertion depth relative to the target. Insertion was performed in 3 to 4 controlled steps. After each advancement, the phantom was returned to the bore for imaging to verify accurate needle progression toward the target.

Fig. 2.

(a) Photo of the biopsy phantom positioned on the MRI bed with the belt securely attached. (b) Photo showing coil placement above the phantom.

The proposed method, from needle path planning to tumor puncture, is summarized in the diagram of Fig. 3.

Fig. 3.

Diagram illustrating the method for calculating the needle guide path and step-by-step tumor puncture.

A T2-W half-Fourier acquisition single-shot turbo spin echo (HASTE) sequence, optimized for minimal scan time while preserving sufficient spatial resolution, was used for syringe alignment with each target. Imaging parameters were: TR/TE = 1500/98 ms, flip angle (FA) = 144°, number of excitations (NEX) = 2, FOV = 346 × 420 mm², slice thickness (ST) = 4 mm, matrix size = 352 × 235, echo train length (ETL) = 81, and pixel bandwidth (pBW) = 676 Hz/pixel, with a total acquisition time of 11s. Quantitative analysis of alignment accuracy involved measuring the offset between the simulated tumor and syringe center on T2-W turbo spin echo (TSE) images (n = 3 per target), referred to as syringe-target (in-plane) alignment accuracy, which was influenced by the baseline accuracy provided by the designed frame.

The same sequence was used to monitor needle advancement. Supplementary imaging employed an ultrafast gradient echo (GRE)-based sequence with a 4 second acquisition time to assess the feasibility of faster imaging in terms of needle visibility and artifact susceptibility. This sequence was a T1-W volumetric interpolated breath-hold examination Dixon (VIBE-Dixon) protocol, acquired with the following parameters: TR/TE = 3.97/ 2.47 ms, pBW = 914 Hz/pixel, FOV = 325 × 400 mm², ST = 3 mm, ETL = 2, matrix = 288 × 199, FA = 9°, and NEX = 1.

Results

Figure 4 displays the results of the planning phase for the central tumor mimic. It includes 2 coronal images: one at the tumor level and the other at the syringe level (Fig. 4a), as well as the fused image (Fig. 4b), which was used to determine the 2D path of the syringe by calculating the required X and Z displacements in mm.

Fig. 4.

(a) Coronal T2-W TSE images acquired before syringe navigation at the level of the targeted tumor mimic and water-filled syringe. (b) Fused image indicating the initial syringe (needle guide) location relative to the targeted tumor mimic.TSE, turbo spin echo; T2-W, T2-weighted.

The accuracy of needle targeting for the centrical tumor is demonstrated in Fig. 5. Figure 5a shows post-navigation T2-W TSE images obtained to verify the alignment of the syringe (and thus the needle guide) with the target. In the axial plane, the syringe direction coincides with the tumor center, which is further confirmed by the fused coronal images displaying the syringe tip centered within the tumor. Figure 5b shows T2-W HASTE images acquired during gradual needle insertion, while Fig. 5c presents corresponding images obtained using the VIBE-Dixon GRE sequence. The metallic needle was visible in both sequences but appeared more sharply defined in the HASTE images. In the VIBE-Dixon images, a pronounced susceptibility artifact was observed around the needle, resulting in an apparent enlargement of its diameter. Additionally, the tumor was not discernible due to insufficient contrast. Therefore, only the HASTE sequence was used for needle guidance toward the second target. Figure 6 demonstrates successful alignment of the needle with the lateral tumor mimic and depicts its stepwise insertion in 4 distinct stages. Across all trials for both tumor mimics, analysis of the deviation from the tumor center confirmed submillimeter in-plane alignment accuracy between the syringe and the target for both simulated tumors.

Fig. 5.

(a) Axial T2-W HASTE image of the phantom following syringe navigation to the central tumor mimic (top), alongside relevant fused coronal images (bottom) showing the alignment of the syringe with the tumor center. (b) Series of axial T2-W HASTE images of the phantom, showing gradual insertion of the biopsy needle towards the central tumor mimic. (c) Corresponding axial T1-W VIBE images illustrating the same stepwise needle insertion.HASTE, half-Fourier acquisition single-shot turbo spin echo; T2-W, T2-weighted; VIBE, volumetric interpolated breath-hold examination.

Fig. 6.

(a) Axial T2-W HASTE image of the phantom following syringe navigation to the lateral tumor mimic (top), alongside relevant fused oblique images (bottom) showing the alignment of the syringe with the tumor. (b) Series of axial T2-W HASTE images of the phantom, showing the gradual insertion of the biopsy needle towards the lateral tumor mimic.HASTE, half-Fourier acquisition single-shot turbo spin echo; T2-W, T2-weighted; VIBE, volumetric interpolated breath-hold examination.

Discussion

This study explored a straightforward ergonomic approach to MRI-guided abdominal biopsy using an MR-compatible X-Z stage biopsy belt, which was subjected to preliminary testing within a high-field conventional MRI scanner. Given its preliminary scope, all experiments were conducted using an agar phantom shaped as a semicircular block to replicate the geometry of the abdomen, featuring high contrast silica-doped tumors.

Needle guide registration in the MRI coordinates was easily achieved using a water-filled syringe as a reference (Fig. 4), as visualized on T2-W HASTE images. Post-navigation imaging (Figs. 5a and 6a) confirmed accurate in-plane needle guide alignment, with an estimated deviation within the submillimeter range. The clinical coaxial biopsy needle was clearly visualized as a region of signal loss while advancing toward each 7 mm tumor (Figs. 5b and 6b), successfully puncturing both the central and lateral targets on the first attempt. Notably, while in-plane needle guide alignment primarily relies on the intrinsic precision of the biopsy frame (defined by its engraved scale), factors such as MRI registration error and imaging resolution can also affect overall accuracy. The study outcomes suggest that, in the present setup, both registration and image resolution were sufficient to enable accurate alignment and reliable targeting.

Susceptibility artifacts from metallic biopsy needles are a well-known challenge in MRI, with their severity depending heavily on the chosen pulse sequence, which must balance imaging speed, resolution, and resistance to artifacts.³⁶ Ultrafast GRE-based protocols (FLASH, VIBE, etc.) are employed in real-time MRI-guided workflows, as they minimize the impact of physiological motion and allow operators to visually track and adjust for respiratory movement during free breathing.³⁷ They are though particularly prone to artifacts such as blooming, signal voids, and spatial distortion, which can obscure anatomical detail or cause displacement of the apparent needle tip. Despite this, their high temporal resolution remains beneficial in procedures where continuous visualization is essential, especially in anatomically complex or motion-sensitive cases.^37–39 In contrast, TSE-based sequences significantly reduce susceptibility artifacts due to their refocusing pulses and reduced T2* sensitivity, yielding better preservation of anatomical detail and reduced spatial distortion.^{36–38,40,41} They thus provide more reliable needle visualization, making them more appropriate for step-by-step guidance,³⁷ where motion can be effectively managed through breath-holding or motion compensation strategies. Material composition also plays a key role. Indicatively, Reichenbach et al.⁴² compared carbon fiber and titanium alloy needles, finding smaller artifacts with carbon fiber, with artifact size amplified on GRE imaging. Beyond sequence and material effects, artifact behavior can also be influenced by needle orientation⁴³ and insertion angle,³⁶ both in terms of artifact size and the accuracy of the needle’s apparent position.

Although a systematic evaluation of sequence types and needle materials was beyond the scope of this study, a clinical coaxial biopsy needle was used to demonstrate basic feasibility and assess artifact behavior with the two employed clinically relevant sequences (HASTE and VIBE-Dixon). While distortion remained within acceptable limits for MR-conditional metallic needles in both sequences, needle visualization was compromised in the GRE-based VIBE sequence due to blooming, and the tumor was not visible, in line with the literature. Overall, these findings support the compatibility of the workflow with standard clinical tools and motivate future preclinical studies to evaluate targeting accuracy across various needle materials and imaging protocols.

Through parameter optimization, the T2-W HASTE sequence was shortened to approximately 10 s while maintaining spatial resolution. This acquisition time is still relatively long compared to durations reported in real-time or ultrafast clinical imaging protocols. Recent advances in MR-guided interventions have enabled ultrafast intraprocedural imaging using advanced pulse sequences, parallel imaging, and non-Cartesian k-space sampling.^37,44–46 These rapid protocols enable real-time feedback and motion compensation, thereby enhancing procedural adaptability and precision. However, they are typically implemented on open MRI systems that permit needle manipulation during imaging. Alternatively, some wide-bore, short-length clinical scanners (such as the 70 cm, 125 cm-long Siemens Espree) allow limited manual access near the bore opening,^44–46 but this is feasible only for certain anatomical regions, such as the anterior abdominal wall or lateral flank. ^37,44–46 Closed-bore systems, in contrast, restrict manual access during imaging, making real-time image-guided interventions in such settings largely dependent on robotic assistance. Otherwise, intermittent imaging guidance (step-by-step technique) is commonly used in clinical practice,^24,47 offering reduced technical complexity, greater flexibility in sequence selection, and broader compatibility across MRI systems.

The system developed in this study falls into the latter category, using an intermittent imaging workflow compatible with standard closed-bore MRI systems. This design prioritizes accessibility and simplicity, making it suitable for institutions without specialized interventional platforms or dedicated real-time imaging systems. The 11s scans were performed during static phases of the procedure and are clinically feasible, especially when paired with breath-holding or motion compensation techniques, as breath-holds of 10–20s are routinely achievable in most patients.⁴⁸

This biopsy device aligns with the observed preference for traditional biopsy techniques among practitioners, despite the advantages of automation and robotics in image-guided interventions.^26,29,30 This preference can be attributed to several factors, including the complexity and cost associated with implementing robotics in MRI,^49–51 as well as a desire to maintain the human element in the process. In this context, the proposed approach is expected to offer advantages over advanced automatic biopsy systems, especially by achieving precise needle guide alignment with a simpler user-friendly setup that requires minimal training, while ensuring human control throughout the process. When compared to conventional free-hand or frame-based methods, this approach is likely to enhance stability and accuracy, thus lowering sampling errors. In addition, it will offer ergonomic benefits that may reduce user fatigue while being cost-efficient for purchasers. Further extensive evaluation is needed to fully assess the advantages and potential limitations of this biopsy technique.

In its current design, the needle guide can be translated along 2 orthogonal axes (X and Z) within the frame to align with the desired entry point, while the frame itself can be repositioned along the belt and secured at various locations around the patient’s circumference, rather than being limited to ventral placement. This design allows multiple surface access points while maintaining procedural simplicity and is expected to perform effectively in abdominal interventions conducted under breath-hold or motion-suppressed conditions,⁴⁸ where minimal target displacement makes a straight access path feasible. However, its fixed needle orientation will limit applicability in procedures requiring angled approaches to account for complex anatomy and residual motion. To broaden its use, future iterations of the device will incorporate a rotatable or articulated needle guide, allowing controlled angular adjustment of the insertion trajectory. This enhancement will enable non-orthogonal targeting and improve access to anatomically challenging locations without requiring fundamental changes to the device’s core design or workflow, which are already compatible with such modifications.

Future studies should include a more extensive phantom evaluation with multiple targeting experiments at varied locations, depths, and insertion angles to assess repeatability and accuracy, refine performance metrics, and support progression to in vivo preclinical testing. Preclinical in vivo studies are essential to fully evaluate the effectiveness and safety of the biopsy belt and procedure, particularly for hard-to-access lesions and sensitive structures and may highlight the need for design modifications or an expanded motion range. In vivo studies should further evaluate respiratory motion effects and implement mitigation strategies such as respiratory gating and advanced motion compensation algorithms, as well as breath-holding techniques in human subjects.^52,53 Importantly, recent advances in respiratory motion compensation, sensor technology, and artificial intelligence offer promising improvements in precision and efficiency for image-guided procedures.⁵³ In this context, the possibility of incorporating motorized motion into the manual biopsy system may also be examined, potentially requiring an expansion of the frame to accommodate piezoelectric motors. In that case, minimizing the size of mechatronic components would be crucial for maintaining the system’s practicality. In optimizing the biopsy belt design and methodology, feedback from future end-users regarding its usability in real-world scenarios will be essential for refining both its design and functionality.

Conclusion

Overall, the proposed biopsy belt and technique may serve as an ergonomic, accessible, and cost-effective alternative to more complex automated systems which typically involve dedicated setups and intricate procedures, necessitating specialized operator training. It has the potential to enhance procedural efficiency and streamline clinician workflow in MRI-guided biopsies, pending comprehensive preclinical validation and confirmation of clinical efficacy. Future indications may include the kidney, pancreas, and liver.