Effects of CT image segmentation methods on the accuracy of long bone 3D reconstructions

An accurate and accessible image segmentation method is in high demand for generating 3D bone models from CT scan data, as such models are required in many areas of medical research. Even though numerous sophisticated segmentation methods have been published over the years, most of them are not readily available to the general research community. Therefore, this study aimed to quantify the accuracy of three popular image segmentation methods, two implementations of intensity thresholding and Canny edge detection, for generating 3D models of long bones. In order to reduce user dependent errors associated with visually selecting a threshold value, we present a new approach of selecting an appropriate threshold value based on the Canny filter. A mechanical contact scanner in conjunction with a microCT scanner was utilised to generate the reference models for validating the 3D bone models generated from CT data of five intact ovine hind limbs. When the overall accuracy of the bone model is considered, the three investigated segmentation methods generated comparable results with mean errors in the range of 0.18-0.24 mm. However, for the bone diaphysis, Canny edge detection and Canny filter based thresholding generated 3D models with a significantly higher accuracy compared to those generated through visually selected thresholds. This study demonstrates that 3D models with sub-voxel accuracy can be generated utilising relatively simple segmentation methods that are available to the general research community.     

Correction of step artefact associated with MRI scanning of long bones

3D models of long bones are being utilised for a number of fields including orthopaedic implant design. Accurate reconstruction of 3D models is of utmost importance to design accurate implants to allow achieving a good alignment between two bone fragments. Thus for this purpose, CT scanners are employed to acquire accurate bone data exposing an individual to a high amount of ionising radiation. Magnetic resonance imaging (MRI) has been shown to be a potential alternative to computed tomography (CT) for scanning of volunteers for 3D reconstruction of long bones, essentially avoiding the high radiation dose from CT. In MRI imaging of long bones, the artefacts due to random movements of the skeletal system create challenges for researchers as they generate inaccuracies in the 3D models generated by using data sets containing such artefacts.One of the defects that have been observed during an initial study is the lateral shift artefact occurring in the reconstructed 3D models. This artefact is believed to result from volunteers moving the leg during two successive scanning stages (the lower limb has to be scanned in at least five stages due to the limited scanning length of the scanner). As this artefact creates inaccuracies in the implants designed using these models, it needs to be corrected before the application of 3D models to implant design. Therefore, this study aimed to correct the lateral shift artefact using 3D modelling techniques.The femora of five ovine hind limbs were scanned with a 3T MRI scanner using a 3D vibe based protocol. The scanning was conducted in two halves, while maintaining a good overlap between them. A lateral shift was generated by moving the limb several millimetres between two scanning stages. The 3D models were reconstructed using a multi threshold segmentation method. The correction of the artefact was achieved by aligning the two halves using the robust iterative closest point (ICP) algorithm, with the help of the overlapping region between the two. The models with the corrected artefact were compared with the reference model generated by CT scanning of the same sample.The results indicate that the correction of the artefact was achieved with an average deviation of 0.32 ± 0.02 mm between the corrected model and the reference model. In comparison, the model obtained from a single MRI scan generated an average error of 0.25 ± 0.02 mm when compared with the reference model. An average deviation of 0.34 ± 0.04 mm was seen when the models generated after the table was moved were compared to the reference models; thus, the movement of the table is also a contributing factor to the motion artefacts.     

Segmentation accuracy of long bones

The use of three-dimensional imaging methodologies in new applications in the orthopaedic field has introduced a need for high accuracy, in addition to a correct diagnosis. The aim of this study was to quantify the absolute dimensional errors between models reconstructed from computed tomography and magnetic resonance images compared to a ground truth for various regions of the bone. Clinical CT and MRI scans were acquired from nine lower leg cadavers and the bones were subsequently cleaned from soft tissues. 3D models of the tibia were created from the segmented CT and MRI images and compared to optical scans of the cleaned bones (considered as ground truth). The 3D reconstruction using CT images resulted in an RMS error of 0.55 mm, corresponding to an overestimated CT bone model compared to the cleaned bone. MR imaging resulted in an RMS error of 0.56 mm; however, the MRI bone model was on average a small underestimation of the cleaned bone. Different regions of the bones were analysed, indicating a difference in accuracy between diaphysis and epiphysis. This study demonstrates a high accuracy for both CT and MRI imaging, supporting the feasibility of using MRI technology for the 3D reconstruction of bones in medical applications.     

基于2D/3D U-plus-net的心脏自动分割

目的:利用2D/3D U-plus-net 提高心脏自动分割的准确率。方法:收集郑州大学第一附属医院60 例患者胸部扫 描CT图像（数据A）及中国科学技术大学附属第一医院45 例患者胸部扫描CT图像（数据B）。基于改进的AlexNet 将 CT 图像分为两类:心脏CT 图像和无心脏CT 图像。在2D/3D U-net 拓扑结构基础上,通过减小网络深度、在长连接中 增加新节点、增加解码器中卷积次数的方法,得到改进后的2D/3D U-plus-net;将靠近腹部的心脏CT图像（图像张数由 预实验决定）输入3D U-plus-net,其余图像输入2D U-plus-net;采用5 倍交叉验证法对模型进行训练及测试。最后通过 Dice 系数、HD95 和平均表面距离（MSD）评估自动分割精度。结果:数据A自动分割的Dice 系数为0.941±0.012,MSD 为（3.918±0.201）mm,HD95为（5.863±0.561）mm;数据B自动分割的Dice系数为0.934±0.014,MSD为（4.112±0.320）mm, HD95 为（6.035±0.659）mm。结论:基于2D/3D U-plus-net 的分割方法提高了心脏自动分割准确率。     

Comparability of skeletal fibulae surfaces generated by different source scanning (dual‐energy CT scan vs. high resolution laser scanning) and 3D geometric morphometric validation

This work aims to test accuracy and comparability of 3D models of human skeletal fibulae generated by clinical CT and laser scanner virtual acquisitions. Mesh topology, segmentation and smoothing protocols were tested to assess variation among meshes generated with different scanning methods and procedures, and to evaluate meshes‐interchangeability in 3D geometric morphometric analysis. A sample of 13 left human fibulae were scanned separately with Revolution Discovery CT dual energy (0.625 mm resolution) and ARTEC Space Spider 3D structured light laser scanner (0.1 mm resolution). Different segmentation methods, including half‐maximum height (HMH) and MIA‐clustering protocols, were compared to their high‐resolution standard generated with laser‐scanner by calculating topological surface deviations. Different smoothing algorithms were also evaluated, such as Laplacian and Taubin smoothing. A total of 142 semilandmarks were used to capture the shape of both proximal and distal fibular epiphyses. After Generalized Procrustes superimposition, the Procrustes coordinates of the proximal and distal fibular epiphyses were used separately to assess variation due to scanning methods and the operator error. Smoothing algorithms at low iteration do not provide significant variation among reconstructions, but segmentation protocol may influence final mesh quality (0.09–0.24 mm). Mean deviation among CT‐generated meshes that were segmented with MIA‐clustering protocol, and laser scanner‐generated ones, is optimal (0.42 mm, ranging 0.35–0.56 mm). Principal component analysis reveals that homologous samples scanned with the two methods cluster together for both the proximal and distal fibular epiphyses. Similarly, Procrustes ANOVA reveals no shape differences between scanning methods and replicates, and only 1.38–1.43% of shape variation is due to scanning device. Topological similarities support the comparability of CT‐ and laser scanner‐generated meshes and validate its simultaneous use in shape analysis with potential clinical relevance. We precautionarily suggest that dedicated trials should be performed in each study when merging different data sources prior to analyses.     

基于级联3D U-Net的CT和MR视交叉自动分割方法

目的:基于级联3D U-Net,利用配对患者头颈部数据［CT和磁共振图像（MRI）］,取得比仅CT数据更高分割精度的视交叉自动分割结果。方法:该级联3D U-Net由一个原始3D U-Net和改进的3D D-S U-Net（3D Deeply-Supervised U-Net）组成,实验使用了60例患者头颈部CT图像及MRI图像（T1和T2模态）,其中随机选取15例患者数据作为测试集,并使用相似性系数（DSC）评估视交叉的自动分割精度。结果:对于测试集中的所有病例,采用多模态数据（CT和MRI）的视交叉的DSC为0.645±0.085,采用单模态数据（CT）的视交叉的DSC为0.552±0.096。结论:基于级联3D U-Net的多模态自动分割模型能够较为准确地实现视交叉的自动分割,且优于仅利用单模态数据的方法,可以辅助医生提高放疗计划制定的工作效率。     

Typical Accuracy and Quality Control of a Process for Creating CT-Based Virtual Bone Models

A pragmatic method for assessing the accuracy and precision of a given processing pipeline required for converting computed tomography (CT) image data of bones into representative three dimensional (3D) models of bone shapes is proposed. The method is based on coprocessing a control object with known geometry which enables the assessment of the quality of resulting 3D models. At three stages of the conversion process, distance measurements were obtained and statistically evaluated. For this study, 31 CT datasets were processed. The final 3D model of the control object contained an average deviation from reference values of −1.07 ± 0.52 mm standard deviation (SD) for edge distances and −0.647 ± 0.43 mm SD for parallel side distances of the control object. Coprocessing a reference object enables the assessment of the accuracy and precision of a given processing pipeline for creating CT-based 3D bone models and is suitable for detecting most systematic or human errors when processing a CT-scan. Typical errors have about the same size as the scan resolution.     

Registration of 2D x-ray images to 3D MRI by generating pseudo-CT data

Spatial and soft tissue information provided by magnetic resonance imaging can be very valuable during image-guided procedures, where usually only real-time two-dimensional (2D) x-ray images are available. Registration of 2D x-ray images to three-dimensional (3D) magnetic resonance imaging (MRI) data, acquired prior to the procedure, can provide optimal information to guide the procedure. However, registering x-ray images to MRI data is not a trivial task because of their fundamental difference in tissue contrast. This paper presents a technique that generates pseudo-computed tomography (CT) data from multi-spectral MRI acquisitions which is sufficiently similar to real CT data to enable registration of x-ray to MRI with comparable accuracy as registration of x-ray to CT. The method is based on a k-nearest-neighbors (kNN)-regression strategy which labels voxels of MRI data with CT Hounsfield Units. The regression method uses multi-spectral MRI intensities and intensity gradients as features to discriminate between various tissue types. The efficacy of using pseudo-CT data for registration of x-ray to MRI was tested on ex vivo animal data. 2D-3D registration experiments using CT and pseudo-CT data of multiple subjects were performed with a commonly used 2D-3D registration algorithm. On average, the median target registration error for registration of two x-ray images to MRI data was approximately 1 mm larger than for x-ray to CT registration. The authors have shown that pseudo-CT data generated from multi-spectral MRI facilitate registration of MRI to x-ray images. From the experiments it could be concluded that the accuracy achieved was comparable to that of registering x-ray images to CT data.     

Dosimetric evaluation of MRI-based treatment planning for prostate cancer

The purpose of this study is to evaluate the dosimetric accuracy of MRI-based treatment planning for prostate cancer using a commercial radiotherapy treatment planning system. Three-dimensional conformal plans for 15 prostate patients were generated using the AcQPlan system. For each patient, dose distributions were calculated using patient CT data with and without heterogeneity correction, and using patient MRI data without heterogeneity correction. MR images were post-processed using the gradient distortion correction (GDC) software. The distortion corrected MR images were fused to the corresponding CT for each patient for target and structure delineation. The femoral heads were delineated based on CT. Other anatomic structures relevant to the treatment (i.e., prostate, seminal vesicles, lymph notes, rectum and bladder) were delineated based on MRI. The external contours were drawn separately on CT and MRI. The same internal contours were used in the dose calculation using CT- and MRI-based geometries by directly transferring them between MRI and CT as needed. Treatment plans were evaluated based on maximum dose, isodose distributions and dose-volume histograms. The results confirm previous investigations that there is no clinically significant dose difference between CT-based prostate plans with and without heterogeneity correction. The difference in the target dose between CT- and MRI-based plans using homogeneous geometry was within 2.5%. Our results suggest that MRI-based treatment planning is suitable for radiotherapy of prostate cancer.     

How does scanner choice and 3D model resolution affect data accuracy?

Researchers using digital methods often collect data from 3D models at different resolutions, obtained using different scanning techniques. Although previous research has sought to understand whether scanning method and model resolution affect data accuracy, no study has systematically evaluated the sources of error associated with scanning method, data acquisition method and model resolution with the aim of providing practical recommendations about the model resolution required to yield sufficiently accurate data for specimens of given sizes. In this study, using data taken from primate specimens of three broad size categories, we test whether 3D models obtained using five different scanners (Breuckmann SmartSCAN, DAVID/HP 3D Pro S3, NextEngine 2020i, Creaform Go!Scan 20 and microCT/clinicalCT) yield accurate measurements. We assess whether caliper measurements can be used alongside measurements collected from 3D surface models, whether scanning resolution affects measurement accuracy, and how scan resolution, estimated using each scanner's proprietary software, compares to model resolution measured in a standardized way. Each scanner produces 3D models that yield accurate measurements for each size category, however, combining caliper data with those taken from digital models can be problematic. Our results indicate that the accuracy of measurements taken from 3D models depends on both object size and model resolution. Based on our findings, we recommend that small specimens should be scanned at <0.3 mm, medium specimens at 0.3–0.7 mm, and large specimens at 0.3–0.5 mm resolutions if data taken from 3D surface models are to be combined with caliper datasets. We further show, for the first time, that discrepancies in estimated final model resolution are frequently observed across software packages. We therefore recommend that researchers ensure that final model resolutions are adequate based on specimen size and are independently verified using a software package other than the scanner's proprietary software. Finally, we consider the implications of the findings that measurements obtained from surface models are variably consistent with those obtained using calipers.     

Accuracy of a Mitral Valve Segmentation Method Using J-Splines for Real-Time 3D Echocardiography Data

Patient-specific models of the heart’s mitral valve (MV) exhibit potential for surgical planning. While advances in 3D echocardiography (3DE) have provided adequate resolution to extract MV leaflet geometry, no study has quantitatively assessed the accuracy of their modeled leaflets vs. a ground-truth standard for temporal frames beyond systolic closure or for differing valvular dysfunctions. The accuracy of a 3DE-based segmentation methodology based on J-splines was assessed for porcine MVs with known 4D leaflet coordinates within a pulsatile simulator during closure, peak closure, and opening for a control, prolapsed, and billowing MV model. For all time points, the mean distance error between the segmented models and ground-truth data were 0.40 ± 0.32 mm, 0.52 ± 0.51 mm, and 0.74 ± 0.69 mm for the control, flail, and billowing models. For all models and temporal frames, 95% of the distance errors were below 1.64 mm. When applied to a patient data set, segmentation was able to confirm a regurgitant orifice and post-operative improvements in coaptation. This study provides an experimental platform for assessing the accuracy of an MV segmentation methodology at phases beyond systolic closure and for differing MV dysfunctions. Results demonstrate the accuracy of a MV segmentation methodology for the development of future surgical planning tools.     

The quality of bone surfaces may govern the use of model based fluoroscopy in the determination of joint laxity

The assessment of knee joint laxity is clinically important but its quantification remains elusive. Calibrated, low dosage fluoroscopy, combined with registered surfaces and controlled external loading may offer possible solutions for quantifying relative tibio-femoral motion without soft tissue artefact, even in native joints. The aim of this study was to determine the accuracy of registration using CT and MRI derived 3D bone models, as well as metallic implants, to 2D single-plane fluoroscopic datasets, to assess their suitability for examining knee joint laxity.Four cadaveric knees and one knee implant were positioned using a micromanipulator. After fluoroscopy, the accuracy of registering each surface to the 2D fluoroscopic images was determined by comparison against known translations from the micromanipulator measurements. Dynamic measurements were also performed to assess the relative tibio-femoral error. For CT and MRI derived 3D femur and tibia models during static testing, the in-plane error was 0.4 mm and 0.9 mm, and out-of-plane error 2.6 mm and 9.3 mm respectively. For metallic implants, the in-plane error was 0.2 mm and out-of-plane error 1.5 mm. The relative tibio-femoral error during dynamic measurements was 0.9 mm, 1.2 mm and 0.7 mm in-plane, and 3.9 mm, 10.4 mm and 2.5 mm out-of-plane for CT and MRI based models and metallic implants respectively. The rotational errors ranged from 0.5° to 1.9° for CT, 0.5–4.3° for MRI and 0.1–0.8° for metallic implants.The results of this study indicate that single-plane fluoroscopic analysis can provide accurate information in the investigation of knee joint laxity, but should be limited to static or quasi-static evaluations when assessing native bones, where possible. With this knowledge of registration accuracy, targeted approaches for the determination of tibio-femoral laxity could now determine objective in vivo measures for the identification of ligament reconstruction candidates as well as improve our understanding of the consequences of knee joint instability in TKA.     

A method to reconstruct patient-specific proximal femur surface models from planar pre-operative radiographs

Three-dimensional reconstruction from volumetric medical images (e.g. CT, MRI) is a well-established technology used in patient-specific modelling. However, there are many cases where only 2D (planar) images may be available, e.g. if radiation dose must be limited or if retrospective data is being used from periods when 3D data was not available. This study aims to address such cases by proposing an automated method to create 3D surface models from planar radiographs. The method consists of (i) contour extraction from the radiograph using an Active Contour (Snake) algorithm, (ii) selection of a closest matching 3D model from a library of generic models, and (iii) warping the selected generic model to improve correlation with the extracted contour. This method proved to be fully automated, rapid and robust on a given set of radiographs. Measured mean surface distance error values were low when comparing models reconstructed from matching pairs of CT scans and planar X-rays (2.57-3.74mm) and within ranges of similar studies. Benefits of the method are that it requires a single radiographic image to perform the surface reconstruction task and it is fully automated. Mechanical simulations of loaded bone with different levels of reconstruction accuracy showed that an error in predicted strain fields grows proportionally to the error level in geometric precision. In conclusion, models generated by the proposed technique are deemed acceptable to perform realistic patient-specific simulations when 3D data sources are unavailable.     

基于人工智能的多模态影像辅助海马体自动勾画研究

目的:利用基于深度学习的人工智能算法,结合头颅MRI和CT的多模态影像,开发海马结构自动勾画技术,为头颅放疗过程中海马体的保护提供高效、准确的自动勾画方法。方法:收集清华大学第一附属医院放疗科从2020年1月~12月就诊的40例脑转移癌患者的定位头颅CT及MRI影像,分别在CT图像、CT-MRI配准图像的两个数据集上训练3D U-Net、3D U-Net Cascade、3D BUC-Net 3个深度学习模型,计算3个模型自动分割的左右海马体与对应的人工标注之间的Dice相似系数（DSC）和95%豪斯多夫距离（95 HD）,以及两者的体积作为模型的分割准确性的评估,并且以对同一大小patch图像的自动分割耗时作为模型效率的评估。结果:引入MRI图像信息对左右海马的自动分割精度有明显的提升;模型3D BUC-Net在CT-MRI数据集上对左右海马体的自动分割都取得最好分割结果（DSC:0.900±0.017,0.882±0.026;95HD:0.792±0.084,0.823±0.093）,而且该模型的分割效率更高。结论:模型3D BUC-Net能在多模态影像上实现高效、准确的海马区的自动勾画,为头颅放疗过程中海马区的保护提供方便。     

Parametric subject-specific model for in vivo 3D reconstruction using bi-planar X-rays: application to the upper femoral extremity

Although feasibility of accurate 3D reconstruction of the proximal epiphysis of the femur from biplanar X-rays (frontal and lateral) has been assessed, in vivo application is limited due to bone superposition. The aim of this study was to propose a specific algorithm to get accurate and reproducible, low dose in vivo 3D reconstruction. To achieve this goal, a parametric subject-specific model was introduced as a priori knowledge. This geometric model was based on a database based on proximal epiphysis of 60 femurs. The accuracy was estimated using comparisons to CT scans on 13 cadaveric femurs, then in vivo intra- and inter- observer reproducibility was assessed using a set of 23 femurs. The mean for the relative difference was 0.2 mm for the in vitro 3D accuracy. The mean error was 1.0 mm with maximum value of 5.1 mm in ideal conditions (in vitro). The confidence interval for the inter-observer reproducibility was within +/-2.2 mm. This method gave us a reproducible tool in order to get in vivo 3D reconstructions of the femur proximal epiphysis from biplanar X-rays.     

三维几何建模与仿真技术在牙颌修复中的应用

提出正常人牙颌组织从CT数据处理到目标组织几何建模及仿真设计的技术路线.将CT断层图像经过线性插值构造出三维数据场,分割出目标牙颌组织后,利用Marching Cubes算法提取牙颌组织的三角面几何模型.采用层切法重建完整的牙列咬和面,其层切精度达到0.2mm.经过配准,实现层切法和CT扫描所得图形线框的坐标拟合.然后在PowerSHAPE中对牙列单元进行三维重建,形成一套易于修改、方便组合的模型库;最后通过编程建立了牙列缺损与修复的几何仿真系统.     

A Semi-automated Approach to Improve the Efficiency of Medical Imaging Segmentation for Haptic Rendering

The Sensimmer platform represents our ongoing research on simultaneous haptics and graphics rendering of 3D models. For simulation of medical and surgical procedures using Sensimmer, 3D models must be obtained from medical imaging data, such as magnetic resonance imaging (MRI) or computed tomography (CT). Image segmentation techniques are used to determine the anatomies of interest from the images. 3D models are obtained from segmentation and their triangle reduction is required for graphics and haptics rendering. This paper focuses on creating 3D models by automating the segmentation of CT images based on the pixel contrast for integrating the interface between Sensimmer and medical imaging devices, using the volumetric approach, Hough transform method, and manual centering method. Hence, automating the process has reduced the segmentation time by 56.35% while maintaining the same accuracy of the output at ±2 voxels.     

Morphometric evaluations of personalised 3D reconstructions and geometric models of the human spine

In the past, several techniques have been developed to study and analyse the 3D characteristics of the human spine: multi-view radiographic or biplanar 3D reconstructions, CT-scan 3D reconstructions and geometric models. Extensive evaluations of three of these techniques that are routinely used at Sainte-Justine Hospital (Montréal, Canada) are presented. The accuracy of these methods is assessed by comparing them with precise measurements made with a coordinate measuring machine on 17 thoracic and lumbar vertebrae (T1-L5) extracted from a normal cadaveric spine specimen. Multi-view radiographic 3D reconstructions are evaluated for different combinations of X-ray views: lateral (LAT), postero-anterior with normal incidence (PAO^o) and postero-anterior with 20^o angled down incidence (PA20^o). The following accuracies are found for these reconstructions obtained from different radiographic setups: 2.1±1.5 mm for the combination with PAO^o-LAT views, and 5.6±4.5 mm for the PAO^o-PA20^o stereopair. Higher errors are found in the postero-anterior direction, especially for the PAO^o-PA20^o view combination. Pedicles are found to be the most precise landmarks. Accuracy for CT-scan 3D reconstructions is about 1.1±0.8 mm. As for a geometric model built using a multiview radiographic reconstruction based on six landmarks per vertebra, accuracies of about 2.6±2.4 mm for landmarks and 2.3±2.0 mm for morphometric parameters are found. The geometric model and 3D reconstruction techniques give accurate information, at low X-ray dose. The accuracy assessment of the techniques used to study the 3D characteristics of the human spine is important, because it allows better and more efficient quantitative evaluations of spinal dysfunctions and their treatments, as well as biomechanical modelling of the spine.     

基于点信息的人体寰椎三维模型配准

目的建立一种基于点信息的寰椎三维模型局部点配准方法,为进行三维数据的统计建模奠定基础。方法以正常人体CT序列图像生成寰椎三维模型30个,所有模型标注人工选择的对应点,设为模板模型1个,训练模型20个,验证模型9个。首先进行训练组模型对模板模型的配准,包括点信息的比较计算和权重系数的机器训练两步,以自动配准点与人工选点的欧式距离之和为测度,获得点配准公式及对应的最佳系数;其次以验证组模型对模板模型进行配准,统计自动配准点与人工选点的欧式距离值,同训练组做对比分析,评估方法的稳定性。结果获得配准函数及对应的最佳权重系数,训练组和验证组配准结果误差分别为1.983和2.045 mm,统计分析表明两组结果没有显著性差异。结论此方法精度及稳定性均达到预期目的,可用于寰椎模型之间感兴趣点的自动配准及统计建模工作中的元素分类。     

Automated localization of implanted seeds in 3D TRUS images used for prostate brachytherapy

An algorithm has been developed in this paper to localize implanted radioactive seeds in 3D ultrasound images for a dynamic intraoperative brachytherapy procedure. Segmentation of the seeds is difficult, due to their small size in relatively low quality of transrectal ultrasound (TRUS) images. In this paper, intraoperative seed segmentation in 3D TRUS images is achieved by performing a subtraction of the image before the needle has been inserted, and the image after the seeds have been implanted. The seeds are searched in a "local" space determined by the needle position and orientation information, which are obtained from a needle segmentation algorithm. To test this approach, 3D TRUS images of the agar and chicken tissue phantoms were obtained. Within these phantoms, dummy seeds were implanted. The seed locations determined by the seed segmentation algorithm were compared with those obtained from a volumetric cone-beam flat-panel micro-CT scanner and human observers. Evaluation of the algorithm showed that the rms error in determining the seed locations using the seed segmentation algorithm was 0.98 mm in agar phantoms and 1.02 mm in chicken phantoms.