The interactive electrode localization utility: software for automatic sorting and labeling of intracranial subdural electrodes

Purpose

Existing methods for sorting, labeling, registering, and across-subject localization of electrodes in intracranial encephalography (iEEG) may involve laborious work requiring manual inspection of radiological images.

Methods

We describe a new open-source software package, the interactive electrode localization utility which presents a full pipeline for the registration, localization, and labeling of iEEG electrodes from CT and MR images. In addition, we describe a method to automatically sort and label electrodes from subdural grids of known geometry.

Results

We validated our software against manual inspection methods in twelve subjects undergoing iEEG for medically intractable epilepsy. Our algorithm for sorting and labeling performed correct identification on 96% of the electrodes.

Conclusions

The sorting and labeling methods we describe offer nearly perfect performance and the software package we have distributed may simplify the process of registering, sorting, labeling, and localizing subdural iEEG grid electrodes by manual inspection.

     

Automatic 3D liver location and segmentation via convolutional neural network and graph cut

Purpose

Segmentation of the liver from abdominal computed tomography (CT) images is an essential step in some computer-assisted clinical interventions, such as surgery planning for living donor liver transplant, radiotherapy and volume measurement. In this work, we develop a deep learning algorithm with graph cut refinement to automatically segment the liver in CT scans.

Methods

The proposed method consists of two main steps: (i) simultaneously liver detection and probabilistic segmentation using 3D convolutional neural network; (ii) accuracy refinement of the initial segmentation with graph cut and the previously learned probability map.

Results

The proposed approach was validated on forty CT volumes taken from two public databases MICCAI-Sliver07 and 3Dircadb1. For the MICCAI-Sliver07 test dataset, the calculated mean ratios of volumetric overlap error (VOE), relative volume difference (RVD), average symmetric surface distance (ASD), root-mean-square symmetric surface distance (RMSD) and maximum symmetric surface distance (MSD) are 5.9, 2.7 %, 0.91, 1.88 and 18.94 mm, respectively. For the 3Dircadb1 dataset, the calculated mean ratios of VOE, RVD, ASD, RMSD and MSD are 9.36, 0.97 %, 1.89, 4.15 and 33.14 mm, respectively.

Conclusions

The proposed method is fully automatic without any user interaction. Quantitative results reveal that the proposed approach is efficient and accurate for hepatic volume estimation in a clinical setup. The high correlation between the automatic and manual references shows that the proposed method can be good enough to replace the time-consuming and nonreproducible manual segmentation method.

     

Automatic labeling of portal and hepatic veins from MR images prior to liver transplantation

Purpose

Living donated liver transplantation is an important task since a person (healthy donor) donates some part of her/his liver to a person in this surgery operation. The success of this operation mainly depends on the sufficiency of vessels and volume of the liver. Accurate labeling of portal and hepatic veins of donors reduces the incidence of complications during and after transplantation. Therefore, prior to the hepatic surgery, automatic analysis and labeling of vasculature structures in the liver are vital to see whether liver is suitable or not for transplantation. However, automatic labeling of veins in the liver is challenging because of partial volume effects, noise and image resolution, which causes wrong connections between vessels. The goal of this paper is to propose an automatic labeling approach for vessels.

Methods

The proposed automated labeling method is based on gray-level values in the MR images and anatomical information. In this work, detection and segmentation of vascular structures in the liver is performed automatically with clustering-based segmentation and refinement stages.

Results

The accuracy of the automatic labeling approach is 85 %. Required processing time for the proposed method (average 6 s) is shorter than manual approach (average 295 s) for labeling of hepatic and portal veins from segmented vessels.

Conclusion

The proposed approach is efficient in terms of both computational cost and accuracy of labeling and segmentation of hepatic and portal veins.

     

3D ultrasound registration-based visual servoing for neurosurgical navigation

Purpose

We present a fully image-based visual servoing framework for neurosurgical navigation and needle guidance. The proposed servo-control scheme allows for compensation of target anatomy movements, maintaining high navigational accuracy over time, and automatic needle guide alignment for accurate manual insertions.

Method

Our system comprises a motorized 3D ultrasound (US) transducer mounted on a robotic arm and equipped with a needle guide. It continuously registers US sweeps in real time with a pre-interventional plan based on CT or MR images and annotations. While a visual control law maintains anatomy visibility and alignment of the needle guide, a force controller is employed for acoustic coupling and tissue pressure. We validate the servoing capabilities of our method on a geometric gel phantom and real human anatomy, and the needle targeting accuracy using CT images on a lumbar spine gel phantom under neurosurgery conditions.

Results

Despite the varying resolution of the acquired 3D sweeps, we achieved direction-independent positioning errors of \(0.35\pm 0.19\) mm and \(0.61^\circ \pm 0.45^\circ \), respectively. Our method is capable of compensating movements of around 25 mm/s and works reliably on human anatomy with errors of \(1.45\pm 0.78\) mm. In all four manual insertions by an expert surgeon, a needle could be successfully inserted into the facet joint, with an estimated targeting accuracy of \(1.33\pm 0.33\) mm, superior to the gold standard.

Conclusion

The experiments demonstrated the feasibility of robotic ultrasound-based navigation and needle guidance for neurosurgical applications such as lumbar spine injections.

     

CT gastrography for volumetric measurement of remnant stomach after distal gastrectomy: a feasibility study

Purpose

To evaluate the feasibility of post-operative CT gastrography for volumetry of the remnant stomach in gastric cancer patients treated with distal gastrectomy.

Methods

CT gastrography was performed with oral administration of effervescent granules in 35 gastric cancer patients who underwent distal gastrectomy. Two readers independently rated the degree of gastric distension on a four-point scale, one (near-total collapse) to four (well distended) and measured the volume of remnant stomach using either 3D or 2D volumetry. The inter-volumetry agreements between the 2D and 3D methods and the interobserver agreements between readers 1 and 2 were assessed by intraclass correlation coefficients (ICCs) and Bland–Altman plots.

Results

The mean score of gastric distension was 3.4 ± 0.6 points and 3.4 ± 0.7 points from readers 1 and 2, respectively. We regarded CT images scored with 3–4 points as a technical success for reliable CT volumetry, which achieved a rate of 91.4% (32/35). For the inter-volumetry agreements between 3D and 2D volumetry, the ICCs were 0.9778 and 0.9814 from readers 1 and 2, respectively. The interobserver agreement between readers 1 and 2 was also excellent, with ICCs of 0.9961 and 0.9876 for 2D and 3D volumetry, respectively. On Bland–Altman plots, the means of differences between any pairs of volumetry measurements ranged from ?31.1 to 3.2 cm³, which may be an acceptable range of measurement variability.

Conclusions

Post-operative CT gastrography is feasible in patients treated with distal gastrectomy. Both 2D and 3D volumetry methods are comparable in measuring the remnant stomach volume.

     

Diffusion-weighted imaging (DWI) of the liver in assessing chronic liver disease: effects of the presence and the degree of ascites on ADC values

Purpose

The aim of this study was to determine the correlation between the liver and spleen apparent diffusion coefficient (ADC) values of patients with chronic liver disease and the presence and the degree of ascites.

Materials and method

In this retrospective study, we assessed 107 patients with chronic liver disease and 39 control subjects who underwent upper abdominal MR imaging including echo-planar diffusion-weighted imaging (DWI). Among the 107 cirrhotic patients, 56 were classified as group 1, 25 as group 2, and 26 as group 3 according to the absence, the presence of minimal, and the presence of massive ascites, respectively. The scores of model for end-stage liver disease (MELD) were matched between groups as the standard reference. The liver ADC, spleen ADC, and normalized liver ADC values were compared between the control group and patients’ groups.

Results

Patients with massive ascites had significantly higher MELD score compared with the other groups. The MELD score was also significantly higher in patient groups than in control group. The liver and normalized liver ADCs of patients’ groups were significantly lower than that of the control group. With some overlap among groups, the measured ADC values decreased as the amount of the ascites increased, and these relationships were statistically significant. Furthermore, compared to control group, patients with massive ascites had significantly higher spleen ADCs.

Conclusion

Our results indicate that the ADC value of the liver and spleen correlates with the presence and the degree of ascites in patients with chronic liver disease, and merits further study.

     

Shape–intensity prior level set combining probabilistic atlas and probability map constrains for automatic liver segmentation from abdominal CT images

Purpose

Propose a fully automatic 3D segmentation framework to segment liver on challenging cases that contain the low contrast of adjacent organs and the presence of pathologies from abdominal CT images.

Methods

First, all of the atlases are weighted in the selected training datasets by calculating the similarities between the atlases and the test image to dynamically generate a subject-specific probabilistic atlas for the test image. The most likely liver region of the test image is further determined based on the generated atlas. A rough segmentation is obtained by a maximum a posteriori classification of probability map, and the final liver segmentation is produced by a shape–intensity prior level set in the most likely liver region. Our method is evaluated and demonstrated on 25 test CT datasets from our partner site, and its results are compared with two state-of-the-art liver segmentation methods. Moreover, our performance results on 10 MICCAI test datasets are submitted to the organizers for comparison with the other automatic algorithms.

Results

Using the 25 test CT datasets, average symmetric surface distance is \(1.09 \pm 0.34\) mm (range 0.62–2.12 mm), root mean square symmetric surface distance error is \(1.72 \pm 0.46\) mm (range 0.97–3.01 mm), and maximum symmetric surface distance error is \(18.04 \pm 3.51\) mm (range 12.73–26.67 mm) by our method. Our method on 10 MICCAI test data sets ranks 10th in all the 47 automatic algorithms on the site as of July 2015. Quantitative results, as well as qualitative comparisons of segmentations, indicate that our method is a promising tool to improve the efficiency of both techniques.

Conclusion

The applicability of the proposed method to some challenging clinical problems and the segmentation of the liver are demonstrated with good results on both quantitative and qualitative experimentations. This study suggests that the proposed framework can be good enough to replace the time-consuming and tedious slice-by-slice manual segmentation approach.

     

Robot-assisted microwave thermoablation of liver tumors: a single-center experience

Purpose

To evaluate and compare the needle placement accuracy, patient dose, procedural time, complication rate and ablation success of microwave thermoablation using a novel robotic guidance approach and a manual approach.

Methods

We performed a retrospective single-center evaluation of 64 microwave thermoablations of liver tumors in 46 patients (10 female, 36 male, mean age 66 years) between June 2014 and February 2015. Thirty ablations were carried out with manual guidance, while 34 ablations were performed using robotic guidance. A 6-week follow-up (ultrasound, computed tomography and MRI) was performed on all patients.

Results

The total procedure time and dose-length product were significantly reduced under robotic guidance (18.3 vs. 21.7 min, \(p<0.001\); 2216 vs. 2881 mGy\(\times \)cm, \(p = 0.04\)). The position of the percutaneous needle was more accurate using robotic guidance (needle deviation 1.6 vs. 3.3 mm, \(p< 0.001\)). There was no significant difference between both groups regarding the complication rate and the ablation success.

Conclusion

Robotic assistance for liver tumor ablation reduces patient dose and allows for fast positioning of the microwave applicator with high accuracy. The complication rate and ablation success of percutaneous microwave thermoablation of malignant liver tumors using either CT fluoroscopy or robotic guidance for needle positioning showed no significant differences in the 6-week follow-up.

     

Automatic abdominal multi-organ segmentation using deep convolutional neural network and time-implicit level sets

Purpose

Multi-organ segmentation from CT images is an essential step for computer-aided diagnosis and surgery planning. However, manual delineation of the organs by radiologists is tedious, time-consuming and poorly reproducible. Therefore, we propose a fully automatic method for the segmentation of multiple organs from three-dimensional abdominal CT images.

Methods

The proposed method employs deep fully convolutional neural networks (CNNs) for organ detection and segmentation, which is further refined by a time-implicit multi-phase evolution method. Firstly, a 3D CNN is trained to automatically localize and delineate the organs of interest with a probability prediction map. The learned probability map provides both subject-specific spatial priors and initialization for subsequent fine segmentation. Then, for the refinement of the multi-organ segmentation, image intensity models, probability priors as well as a disjoint region constraint are incorporated into an unified energy functional. Finally, a novel time-implicit multi-phase level-set algorithm is utilized to efficiently optimize the proposed energy functional model.

Results

Our method has been evaluated on 140 abdominal CT scans for the segmentation of four organs (liver, spleen and both kidneys). With respect to the ground truth, average Dice overlap ratios for the liver, spleen and both kidneys are 96.0, 94.2 and 95.4%, respectively, and average symmetric surface distance is less than 1.3 mm for all the segmented organs. The computation time for a CT volume is 125 s in average. The achieved accuracy compares well to state-of-the-art methods with much higher efficiency.

Conclusion

A fully automatic method for multi-organ segmentation from abdominal CT images was developed and evaluated. The results demonstrated its potential in clinical usage with high effectiveness, robustness and efficiency.

     

Olanzapine vs haloperidol: treating delirium in a critical care setting

Objective

To compare the safety and estimate the response profile of olanzapine, a second-generation antipsychotic, to haloperidol in the treatment of delirium in the critical care setting.

Design

Prospective randomized trial

Setting

Tertiary care university affiliated critical care unit.

Patients

All admissions to a medical and surgical intensive care unit with a diagnosis of delirium.

Interventions

Patients were randomized to receive either enteral olanzapine or haloperidol.

Measurements

Patient’s delirium severity and benzodiazepine use were monitored over 5 days after the diagnosis of delirium.

Main results

Delirium Index decreased over time in both groups, as did the administered dose of benzodiazepines. Clinical improvement was similar in both treatment arms. No side effects were noted in the olanzapine group, whereas the use of haloperidol was associated with extrapyramidal side effects.

Conclusions

Olanzapine is a safe alternative to haloperidol in delirious critical care patients, and may be of particular interest in patients in whom haloperidol is contraindicated.

     

A magnetic resonance image-guided breast needle intervention robot system: overview and design considerations

Purpose

We developed an image-guided intervention robot system that can be operated in a magnetic resonance (MR) imaging gantry. The system incorporates a bendable needle intervention robot for breast cancer patients that overcomes the space limitations of the MR gantry.

Methods

Most breast coil designs for breast MR imaging have side openings to allow manual localization. However, for many intervention procedures, the patient must be removed from the gantry. A robotic manipulation system with integrated image guidance software was developed. Our robotic manipulator was designed to be slim, so as to fit between the patient’s side and the MR gantry wall. Only non-magnetic materials were used, and an electromagnetic shield was employed for cables and circuits. The image guidance software was built using open source libraries. In situ feasibility tests were performed in a 3-T MR system. One target point in the breast phantom was chosen by the clinician for each experiment, and our robot moved the needle close to the target point.

Results

Without image-guided feedback control, the needle end could not hit the target point (distance = 5 mm) in the first experiment. Using our robotic system, the needle hits the target lesion of the breast phantom at a distance of 2.3 mm from the same target point using image-guided feedback. The second experiment was performed using other target points, and the distance between the final needle end point and the target point was 0.8 mm.

Conclusions

We successfully developed an MR-guided needle intervention robot for breast cancer patients. Further research will allow the expansion of these interventions.

     

Multi-ray medical ultrasound simulation without explicit speckle modelling

Purpose

To develop a medical ultrasound (US) simulation method using T1-weighted magnetic resonance images (MRI) as the input that offers a compromise between low-cost ray-based and high-cost realistic wave-based simulations.

Methods

The proposed method uses a novel multi-ray image formation approach with a virtual phased array transducer probe. A domain model is built from input MR images. Multiple virtual acoustic rays are emerged from each element of the linear transducer array. Reflected and transmitted acoustic energy at discrete points along each ray is computed independently. Simulated US images are computed by fusion of the reflected energy along multiple rays from multiple transducers, while phase delays due to differences in distances to transducers are taken into account. A preliminary implementation using GPUs is presented.

Results

Preliminary results show that the multi-ray approach is capable of generating view point-dependent realistic US images with an inherent Rician distributed speckle pattern automatically. The proposed simulator can reproduce the shadowing artefacts and demonstrates frequency dependence apt for practical training purposes. We also have presented preliminary results towards the utilization of the method for real-time simulations.

Conclusions

The proposed method offers a low-cost near-real-time wave-like simulation of realistic US images from input MR data. It can further be improved to cover the pathological findings using an improved domain model, without any algorithmic updates. Such a domain model would require lesion segmentation or manual embedding of virtual pathologies for training purposes.

     

Automated liver segmentation from a postmortem CT scan based on a statistical shape model

Purpose

Automated liver segmentation from a postmortem computed tomography (PMCT) volume is a challenging problem owing to the large deformation and intensity changes caused by severe pathology and/or postmortem changes. This paper addresses this problem by a novel segmentation algorithm using a statistical shape model (SSM) for a postmortem liver.

Methods

The location and shape parameters of a liver are directly estimated from a given volume by the proposed SSM-guided expectation–maximization (EM) algorithm without any spatial standardization that might fail owing to the large deformation and intensity changes. The estimated location and shape parameters are then used as a constraint of the subsequent fine segmentation process based on graph cuts. Algorithms with eight different SSMs were trained using 144 in vivo and 32 postmortem livers, and the segmentation algorithm was tested on 32 postmortem livers in a twofold cross validation manner. The segmentation performance is measured by the Jaccard index (JI) between the segmentation result and the true liver label.

Results

The average JI of the segmentation result with the best SSM was 0.8501, which was better compared with the results obtained using conventional SSMs and the results of the previous postmortem liver segmentation with statistically significant difference.

Conclusions

We proposed an algorithm for automated liver segmentation from a PMCT volume, in which an SSM-guided EM algorithm estimated the location and shape parameters of a liver in a given volume accurately. We demonstrated the effectiveness of the proposed algorithm using actual postmortem CT volumes.

     

Orbital volume analysis: validation of a semi-automatic software segmentation method

Purpose

The purpose of this study was to validate a quick, accurate and reproducible (semi-) automatic software segmentation method to measure orbital volume in the unaffected bony orbit. Precise volume measurement of the orbital cavity is a useful addition to pre-operative planning and intraoperative navigation in orbital reconstruction.

Methods

In 21 CT scans, one unaffected orbit was selected to compare manual segmentation (gold standard) with three segmentation methods using iPlan software (version 3.0.5; Brainlab, Feldkirchen, Germany): automatic (method A), automatic minus bone/air masks (method SA) and automatic minus masks followed by manual adjustments (method SAA). First, validation of the manual segmentation and a newly described method for the anterior boundary was performed. Subsequently the accuracy, reproducibility and time efficiency of the methods were examined. Measurements were performed by two observers.

Results

The intraclass correlation for the interobserver agreement of the anterior boundary was 0.992, and the intraobserver and interobserver agreement for the manual segmentation were 0.997 and 0.994, respectively. Method A had an average volumetric difference of 0.49 cc (SD 0.74) in comparison with the gold standard; this was 0.24 cc (SD 0.27) for method SA and 0.86 cc (SD 0.27) for method SAA. The average time for each method was 38 (SD 5.4), 146 (SD 16.0) and 327 (SD 36.2) seconds per orbit.

Conclusion

The built-in automatic method A is quick, but suboptimal for clinical use. The newly developed method SA appears to be accurate, reproducible, quick and easy to use. Manual adjustments in method SAA are more time-consuming and do not improve volume accuracy. The largest volume discrepancy is located near the inferior orbital fissure.

     

Manuelle Triggerpunkttherapie bei Schulterschmerzen

Background

Although chronic shoulder pain is highly prevalent and myofascial trigger points (mTrP) are thought to be found in the majority of patients with shoulder complaints, the influence on the pain mechanism remains unclear. There are only very few controlled clinical studies on the effects of manual trigger point compression therapy.

Objective

This randomized controlled trial (RCT) compared the short-term effects of manual trigger point compression therapy (n = 6) with manual sham therapy (n = 6) in patients with unilateral shoulder pain due to myofascial syndrome (MFS).

Material and methods

The measurement data were collected before and after two sessions of therapy. Pressure pain thresholds (PPT) of mTrP and symmetrically located points on the asymptomatic side were measured together with neutral points in order to detect a potential unilateral or generalized hyperalgesia. Additionally, the pain was assessed on a visual analog scale (VAS) at rest and during movement and the neck disability index (NDI) and disabilities of the arm, shoulder and hand (DASH) questionnaires were also completed and evaluated.

Results

Both treatment modalities led to a significant improvement; however, the manual trigger point compression therapy was significantly more effective in comparison to sham therapy, as measured by different parameters.

Conclusion

The significant improvement of PPT values in the interventional group even at sites that were not directly treated, indicates central mechanisms in pain threshold modulation induced by manual compression therapy. The weaker but still measurable effects of sham therapy might be explained by the sham modality being a hands on technique or by sufficient stimulation of the trigger point region during the diagnostics and PPT measurements.

     

Automated muscle segmentation from CT images of the hip and thigh using a hierarchical multi-atlas method

Purpose

Patient-specific quantitative assessments of muscle mass and biomechanical musculoskeletal simulations require segmentation of the muscles from medical images. The objective of this work is to automate muscle segmentation from CT data of the hip and thigh.

Method

We propose a hierarchical multi-atlas method in which each hierarchy includes spatial normalization using simpler pre-segmented structures in order to reduce the inter-patient variability of more complex target structures.

Results

The proposed hierarchical method was evaluated with 19 muscles from 20 CT images of the hip and thigh using the manual segmentation by expert orthopedic surgeons as ground truth. The average symmetric surface distance was significantly reduced in the proposed method (1.53 mm) in comparison with the conventional method (2.65 mm).

Conclusion

We demonstrated that the proposed hierarchical multi-atlas method improved the accuracy of muscle segmentation from CT images, in which large inter-patient variability and insufficient contrast were involved.

     

Vessel-based registration of an optical shape sensing catheter for MR navigation

Purpose

Magnetic resonance navigation (MRN), achieved with an upgraded MRI scanner, aims to guide therapeutic nanoparticles from their release in the hepatic vascular network to embolize highly vascularized liver tumors. Visualizing the catheter in real-time within the arterial network is important for selective embolization within the MR gantry. To achieve this, a new MR-compatible catheter tracking technology based on optical shape sensing is used.

Methods

This paper proposes a vessel-based registration pipeline to co-align this novel catheter tracking technology to the patient’s diagnostic MR angiography (MRA) with 3D roadmapping. The method first extracts the 3D hepatic arteries from a diagnostic MRA based on concurrent deformable models, creating a detailed representation of the patient’s internal anatomy. Once the optical shape sensing fibers, inserted in a double-lumen catheter, is guided into the hepatic arteries, the 3D centerline of the catheter is inferred and updated in real-time using strain measurements derived from fiber Bragg gratings sensors. Using both centerlines, a diffeomorphic registration based on a spectral representation of the high-level geometrical primitives is applied.

Results

Results show promise in registration accuracy in five phantom models created from stereolithography of patient-specific vascular anatomies, with maximum target registration errors below 2 mm. Furthermore, registration accuracy with the shape sensing tracking technology remains insensitive to the magnetic field of the MR magnet.

Conclusions

This study demonstrates that an accurate registration procedure of a shape sensing catheter with diagnostic imaging is feasible.

     

Hepatic imaging following intra-arterial embolotherapy

Purpose

To discuss guidelines and salient imaging findings of solid tumors treated with common intra-arterial procedures used in interventional oncology.

Methods

A meticulous literature search of PubMed-indexed articles was conducted. Key words included “imaging + embolization,” “imaging + TACE,” “imaging + radioembolization,” “imaging + Y90,” “mRECIST,” and “EASL.” Representative post-treatment cross-sectional images were obtained from past cases in this institution.

Results

Intra-arterial therapy (IAT) in interventional oncology includes bland embolization, chemoembolization, and radioembolization. Solid tumors of the liver are the primary focus of these procedures. Cross-sectional CT and/or MR are the main modalities used to image tumors after treatment. Traditional size-based response criteria (WHO and RECIST) alone are of limited utility in determining response to IAT; tumoral necrosis and enhancement must be considered. Specifically for HCC, the EASL and mRECIST guidelines are becoming widely adopted response criteria to assess these factors. DWI, FDG-PET, and CEUS are modalities that play an adjunctive but controversial role.

Conclusions

Radiologists must be aware that the different forms of intra-arterial therapy yield characteristic findings on cross-sectional imaging. Knowledge of these findings is integral to accurate assessment of tumor response and progression.

     

Landmark-guided diffeomorphic demons algorithm and its application to automatic segmentation of the whole spine and pelvis in CT images

Purpose

A fully automatic multiatlas-based method for segmentation of the spine and pelvis in a torso CT volume is proposed. A novel landmark-guided diffeomorphic demons algorithm is used to register a given CT image to multiple atlas volumes. This algorithm can utilize both grayscale image information and given landmark coordinate information optimally.

Methods

The segmentation has four steps. Firstly, 170 bony landmarks are detected in the given volume. Using these landmark positions, an atlas selection procedure is performed to reduce the computational cost of the following registration. Then the chosen atlas volumes are registered to the given CT image. Finally, voxelwise label voting is performed to determine the final segmentation result.

Results

The proposed method was evaluated using 50 torso CT datasets as well as the public SpineWeb dataset. As a result, a mean distance error of \(0.59\pm 0.14\hbox { mm}\) and a mean Dice coefficient of \(0.90\pm 0.02\) were achieved for the whole spine and the pelvic bones, which are competitive with other state-of-the-art methods.

Conclusion

From the experimental results, the usefulness of the proposed segmentation method was validated.