Interventional radiologists have long faced a stubborn dilemma at the operating table: how to confirm that a biopsy needle has landed exactly where it should without blasting the patient with yet another scan. A new study from Memorial Sloan Kettering Cancer Center offers a striking answer. By using software that virtually reconstructs the position of a needle inside a pre-existing cone-beam CT scan, drawn from just two fluoroscopic images, the researchers achieved a median discrepancy of only 1.3 millimeters between the predicted and actual needle tip positions. The work, published in CVIR Oncology, evaluated 40 procedures in cancer patients and found that every single virtual reconstruction correctly mirrored the safety profile of the real needle path, avoiding all critical structures such as major arteries, veins, and nerves. No procedural adverse events were recorded across the entire cohort, and all 40 procedures achieved their technical goals.
The technology at the heart of the study, known as needle virtual reconstruction or NVR, addresses a fundamental bottleneck in image-guided interventions. Percutaneous procedures, in which needles are advanced through the skin to reach deep targets such as tumors or bone lesions, are commonly performed under cone-beam CT guidance. Cone-beam CT holds significant advantages over conventional CT in the interventional suite: it avoids the constraints of bore size and gantry orientation, and it integrates seamlessly with real-time fluoroscopy, allowing clinicians to visualize instruments continuously during an intervention. But there is a catch. Most commercially available guidance systems allow operators to plan a needle trajectory on an initial CBCT acquisition and monitor progress using fluoroscopy overlays, yet evaluating the needle’s latest position within the three-dimensional volume requires acquiring an entirely new CBCT scan. Each additional acquisition adds seconds to minutes of procedure time and delivers another dose of ionizing radiation to both patient and operator.
NVR software, marketed as Needle ASSIST with Stereo 3D by GE HealthCare, sidesteps this problem by mathematical triangulation. Instead of a full volumetric rotation, the system needs only two fluoroscopic projections captured with the needle in place. Because the needle’s silhouette appears from two different angles, and because the geometry of the fluoroscopic gantry is precisely known, the software can reconstruct the needle’s three-dimensional position and map it into the coordinate space of the initial planning CBCT. The virtual needle then appears alongside the planned trajectory, giving the operator an immediate, radiation-light assessment of whether the instrument is on course, short of target, or straying toward danger. The workflow relies on two automatically generated views: a so-called bull’s eye view that looks directly down the barrel of the needle to guide entry point and orientation, and a progress view that displays the full needle shaft to track insertion depth.
To validate this approach rigorously, the research team conducted a retrospective single-center cohort study at a large cancer hospital, with institutional review board approval under protocol 16-402. They enrolled consecutive patients who underwent percutaneous image-guided interventions using NVR software between January 2023 and November 2024. The final cohort comprised 40 procedures in 40 cancer patients: 37 biopsies and 3 kyphoplasties, the latter being vertebral augmentation procedures in which cement is delivered into collapsed vertebrae. All spinal interventions followed a transpedicular approach, threading the needle through the bony pedicle of the vertebra, while other procedures used anatomically safe paths selected to avoid critical structures. Every case followed societal guidelines for anticoagulation management and bleeding risk, and procedures were performed under monitored anesthesia care or general anesthesia by an interventional radiologist with more than two decades of experience.
The imaging protocol followed a standardized sequence. Before any needle was inserted, an initial cone-beam CT was acquired using a 200-degree rotational scan at 40 degrees per second over roughly five seconds, capturing between 147 and 244 frames depending on the suite. Variable kVp and mAs settings dynamically compensated for differences in patient anatomy, and the reconstructed three-dimensional volume spanned a 24-centimeter-diameter cylinder in a 512-cubed matrix. The trajectory was planned tableside by defining target and entry points, and the planned line was overlaid on live fluoroscopy. Bone anatomy derived from the initial CBCT was superimposed to confirm accurate registration and flag any patient movement, with tableside adjustments made as needed. After needle insertion, the virtual reconstruction was generated automatically from two fluoroscopic projections, and critically, a final CBCT was acquired as ground truth to validate needle position before the intervention proceeded. It is this final scan that allowed the team to measure, in hindsight, how well the virtual needle matched reality.
The validation methodology was deliberately conservative. The researchers registered the pre-insertion planning CBCT with the final ground-truth CBCT based on bony landmarks, without displaying the virtual reconstruction. They then measured the maximal distance between the virtual needle tip and the actual needle tip on the bull’s eye view using a dedicated imaging workstation. Two authors performed the measurements independently and reached consensus, with a third author arbitrating disagreements. In parallel, they assessed safety agreement: reviewers first confirmed on the final CBCT that all real needle trajectories avoided critical structures, then checked whether the software’s virtual predictions had correctly indicated those same safe paths. The results were unambiguous. The median discrepancy between virtual and actual needle tip positions was 1.3 millimeters, with an interquartile range of 0.9 to 2.2 millimeters, and the virtual reconstructions correctly predicted the safety profile in every case, with no erroneous depiction of a needle touching a critical structure.
Procedural metrics from the study paint a picture of efficient, low-dose practice. The median procedure duration was 57.5 minutes, median fluoroscopy time was 128 seconds, and the median total dose area product, combining fluoroscopy and CBCT, was 28.9 Gy·cm². Technical success, defined as placement of the needle tip within the boundaries of the targeted lesion or bone on the final CBCT, was achieved in 100 percent of cases. For the applicable tumor cases, the median target lesion size was 24 millimeters. The cohort itself skewed older and heavier, with a median age of 71 years and a median body mass index of 26.1, a detail that matters because larger patients typically require higher radiation doses for adequate image quality. Despite this, the reported doses remained within published reference levels for comparable procedures, suggesting that the guidance software did not inflate radiation burden even in technically challenging anatomy.
The broader implications reach beyond the operating suite. Previous research has suggested that virtual needle guidance technologies can meaningfully cut radiation exposure by eliminating the multiple verification CBCT scans traditionally interspersed through needle positioning. One earlier study cited by the authors reported reductions in air kerma and dose area product of 27 percent and 35 percent respectively when such technology replaced repeated volumetric checks. The accuracy figures now reported go further, raising the possibility that even the final confirmation CBCT, acquired after every needle placement in this study, might eventually be dispensable for select cases, with two quick fluoroscopic views providing sufficient reassurance more simply and faster. The authors are careful on this point: eliminating the final CBCT would require further evaluation, including a randomized or controlled comparison, before it could become standard practice. Guidance software of this kind has already found roles in musculoskeletal intervention, endoleak treatment, and thermal ablation, but the new study is distinctive in specifically quantifying reconstruction accuracy against a ground-truth scan.
The study is not without caveats, and the authors lay them out candidly. It was a single-center retrospective analysis spanning a variety of anatomical targets, so multi-institutional validation would strengthen the findings. The absence of a control group means the team cannot definitively attribute reductions in radiation dose or procedure time to the software itself. Accuracy was measured in a two-dimensional plane on the bull’s eye view rather than in full three dimensions, which would offer a more complete picture of reconstruction error. And the operating physician’s substantial experience, more than twenty years of practice, may have influenced fluoroscopy times and radiation doses in ways that might not generalize to less seasoned operators. There is also a commercial relationship to note: two of the authors serve as consultants for GE HealthCare, the manufacturer of the software and imaging systems used in the study, and the research received support in part through a National Institutes of Health cancer center support grant.
Even with those limitations, the study marks a persuasive step toward smarter, leaner image-guided surgery. The vision it sketches is an interventional suite in which a single volumetric scan at the start of a procedure anchors all subsequent navigation, and every subsequent question about needle position is answered with two milliseconds-fast fluoroscopic frames rather than a full rotation of the C-arm. For cancer patients, who often undergo repeated biopsies and ablative procedures over the course of their illness, the cumulative savings in radiation, anesthesia time, and waiting could be substantial. For the clinicians, a trustworthy virtual needle means fewer interruptions, faster workflows, and continuous three-dimensional awareness of where their instrument truly sits relative to vessels, nerves, and bone. If larger controlled studies confirm these results, the humble biopsy needle may soon travel through the body shadowed by a digital twin that is accurate to little more than a millimeter, and the era of confirming positions with repeated scans may give way to one of calculating them.
Subject of Research: Accuracy and safety of virtual needle reconstruction software during cone-beam CT-guided percutaneous procedures in cancer patients
Article Title: Virtual reconstruction to assess needle position during percutaneous procedures performed under cone-beam computed tomography: safety and accuracy
Article References: Geevarghese, R., Kiely, L., Petre, E. N., Solomon, S. B., & Cornelis, F. H. (2026). Virtual reconstruction to assess needle position during percutaneous procedures performed under cone-beam computed tomography: safety and accuracy. CVIR Oncology, 2(1), Article 9. https://doi.org/10.1007/s44343-026-00042-6
Image Credits: AI Generated
DOI: 10.1007/s44343-026-00042-6
Keywords: cone-beam CT, needle virtual reconstruction, interventional radiology, percutaneous biopsy, image-guided procedures, fluoroscopy, radiation dose, cancer, technical success, kyphoplasty, needle guidance, medical imaging
Cite Scienmag News
APA MLA Chicago
Nathaniel Bowman. (September 23, 2026). Virtual Needle Reconstruction With Cone-Beam CT Hits 1.3 Millimeter Accuracy in Cancer Procedures. Scienmag. https://scienmag.com/virtual-needle-reconstruction-with-cone-beam-ct-hits-1-3-millimeter-accuracy-in-cancer-procedures/
Nathaniel Bowman. “Virtual Needle Reconstruction With Cone-Beam CT Hits 1.3 Millimeter Accuracy in Cancer Procedures.” Scienmag, 23 September 2026, https://scienmag.com/virtual-needle-reconstruction-with-cone-beam-ct-hits-1-3-millimeter-accuracy-in-cancer-procedures/. Accessed 23 September 2026.
Nathaniel Bowman. “Virtual Needle Reconstruction With Cone-Beam CT Hits 1.3 Millimeter Accuracy in Cancer Procedures.” Scienmag. September 23, 2026. https://scienmag.com/virtual-needle-reconstruction-with-cone-beam-ct-hits-1-3-millimeter-accuracy-in-cancer-procedures/
Copy citation Download RIS
Tags: advancements in cancer procedural imagingcancercone-beam CTcone-beam CT guided cancer biopsiescone-beam CT versus traditional imagingfluoroscopyimage-guided percutaneous interventionsimage-guided proceduresinterventional radiologyinterventional radiology needle placement accuracykyphoplastyMedical Imagingminimally invasive tumor biopsy techniquesneedle guidanceneedle virtual reconstructionNVR technology in cancer procedurespercutaneous biopsyprecision in tumor targeting with cone-beam CTradiation dosereal-time imaging in interventional radiologyreducing radiation exposure during biopsiessafety assessment of needle trajectoriestechnical successvirtual needle reconstruction
