Jef Vandemeulebroucke

Evaluation of Registration Methods on Thoracic CT: The EMPIRE10 Challenge

EMPIRE10 (Evaluation of Methods for Pulmonary Image REgistration 2010) is a public platform for fair and meaningful comparison of registration algorithms which are applied to a database of intra patient thoracic CT image pairs. Evaluation of nonrigid registration techniques is a nontrivial task. This is compounded by the fact that researchers typically test only on their own data, which varies widely. For this reason, reliable assessment and comparison of different registration algorithms has been virtually impossible in the past. In this work we present the results of the launch phase of EMPIRE10, which comprised the comprehensive evaluation and comparison of 20 individual algorithms from leading academic and industrial research groups. All algorithms are applied to the same set of 30 thoracic CT pairs. Algorithm settings and parameters are chosen by researchers expert in the con figuration of their own method and the evaluation is independent, using the same criteria for all participants. All results are published on the EMPIRE10 website (http://empire10.isi.uu.nl). The challenge remains ongoing and open to new participants. Full results from 24 algorithms have been published at the time of writing. This paper details the organization of the challenge, the data and evaluation methods and the outcome of the initial launch with 20 algorithms. The gain in knowledge and future work are discussed.

Spatiotemporal motion estimation for respiratory-correlated imaging of the lungs

PURPOSE Four-dimensional computed tomography (4D CT) can provide patient-specific motion information for radiotherapy planning and delivery. Motion estimation in 4D CT is challenging due to the reduced image quality and the presence of artifacts. We aim to improve the robustness of deformable registration applied to respiratory-correlated imaging of the lungs, by using a global problem formulation and pursuing a restrictive parametrization for the spatiotemporal deformation model. METHODS A spatial transformation based on free-form deformations was extended to the temporal domain, by explicitly modeling the trajectory using a cyclic temporal model based on B-splines. A global registration criterion allowed to consider the entire image sequence simultaneously and enforce the temporal coherence of the deformation throughout the respiratory cycle. To ensure a parametrization capable of capturing the dynamics of respiratory motion, a prestudy was performed on the temporal dimension separately. The temporal parameters were tuned by fitting them to diaphragm motion data acquired for a large patient group. Suitable properties were retained and applied to spatiotemporal registration of 4D CT data. Registration results were validated using large sets of landmarks and compared to consecutive spatial registrations. To illustrate the benefit of the spatiotemporal approach, we also assessed the performance in the presence of motion-induced artifacts. RESULTS Cubic B-splines gave better or similar fitting results as lower orders and were selected because of their inherently stronger regularization. The fitting and registration errors increased gradually with the temporal control point spacing, representing a trade-off between achievable accuracy and sensitivity to noise and artifacts. A piecewise smooth trajectory model, allowing for a discontinuous change of speed at end-inhale, was found most suitable to account for the sudden changes of motion at this breathing phase. The spatiotemporal modeling allowed a reduction of the number of parameters of 45%, while maintaining registration accuracy within 0.1 mm. The approach reduced the sensitivity to artifacts. CONCLUSIONS Spatiotemporal registration can provide accurate motion estimation for 4D CT and improves the robustness to artifacts.

Automated segmentation of a motion mask to preserve sliding motion in deformable registration of thoracic CT

PURPOSE Deformable registration generally relies on the assumption that the sought spatial transformation is smooth. Yet, breathing motion involves sliding of the lung with respect to the chest wall, causing a discontinuity in the motion field, and the smoothness assumption can lead to poor matching accuracy. In response, alternative registration methods have been proposed, several of which rely on prior segmentations. We propose an original method for automatically extracting a particular segmentation, called a motion mask, from a CT image of the thorax. METHODS The motion mask separates moving from less-moving regions, conveniently allowing simultaneous estimation of their motion, while providing an interface where sliding occurs. The sought segmentation is subanatomical and based on physiological considerations, rather than organ boundaries. We therefore first extract clear anatomical features from the image, with respect to which the mask is defined. Level sets are then used to obtain smooth surfaces interpolating these features. The resulting procedure comes down to a monitored level set segmentation of binary label images. The method was applied to sixteen inhale-exhale image pairs. To illustrate the suitability of the motion masks, they were used during deformable registration of the thorax. RESULTS For all patients, the obtained motion masks complied with the physiological requirements and were consistent with respect to patient anatomy between inhale and exhale. Registration using the motion mask resulted in higher matching accuracy for all patients, and the improvement was statistically significant. Registration performance was comparable to that obtained using lung masks when considering the entire lung region, but the use of motion masks led to significantly better matching near the diaphragm and mediastinum, for the bony anatomy and for the trachea. The use of the masks was shown to facilitate the registration, allowing to reduce the complexity of the spatial transformation considerably, while maintaining matching accuracy. CONCLUSIONS We proposed an automated segmentation method for obtaining motion masks, capable of facilitating deformable registration of the thorax. The use of motion masks during registration leads to matching accuracies comparable to the use of lung masks for the lung region but motion masks are more suitable when registering the entire thorax.

Is abdominal compression useful in lung stereotactic body radiation therapy? A 4DCT and dosimetric lobe-dependent study

PURPOSE To determine the usefulness of abdominal compression in lung stereotactic body radiation therapy (SBRT) depending on lobe tumor location. MATERIALS AND METHODS Twenty-seven non-small cell lung cancer patients were immobilized in the Stereotactic Body Frame™ (Elekta). Eighteen tumors were located in an upper lobe, one in the middle lobe and nine in a lower lobe (one patient had two lesions). All patients underwent two four-dimensional computed tomography (4DCT) scans, with and without abdominal compression. Three-dimensional tumor motion amplitude was determined using manual landmark annotation. We also determined the internal target volume (ITV) and the influence of abdominal compression on lung dose-volume histograms. RESULTS The mean reduction of tumor motion amplitude was 3.5 mm (p = 0.009) for lower lobe tumors and 0.8 mm (p = 0.026) for upper/middle lobe locations. Compression increased tumor motion in 5 cases. Mean ITV reduction was 3.6 cm(3) (p = 0.039) for lower lobe and 0.2 cm(3) (p = 0.048) for upper/middle lobe lesions. Dosimetric gain of the compression for lung sparing was not clinically relevant. CONCLUSIONS The most significant impact of abdominal compression was obtained in patients with lower lobe tumors. However, minor or negative effects of compression were reported for other patients and lung sparing was not substantially improved. At our institute, patients with upper or middle lobe lesions are now systematically treated without compression and the usefulness of compression for lower lobe tumors is evaluated on an individual basis.

Respiratory Motion Estimation from Cone-Beam Projections Using a Prior Model

Respiratory motion introduces uncertainties when planning and delivering radiotherapy for lung cancer patients. Cone-beam projections acquired in the treatment room could provide valuable information for building motion models, useful for gated treatment delivery or motion compensated reconstruction. We propose a method for estimating 3D+T respiratory motion from the 2D+T cone-beam projection sequence by including prior knowledge about the patients breathing motion. Motion estimation is accomplished by maximizing the similarity of the projected view of a patient specific model to observed projections of the cone-beam sequence. This is done semi-globally, considering entire breathing cycles. Using realistic patient data, we show that the method is capable of good prediction of the internal patient motion from cone-beam data, even when confronted with interfractional changes in the breathing motion.

Predictive Value of Pattern Classification 24 Hours after Radiofrequency Ablation of Liver Metastases on CT and Positron Emission Tomography/CT

PURPOSE To assess a classification scheme for predicting local tumor progression (LTP) after radiofrequency (RF) ablation of liver metastases, using predefined patterns on contrast-enhanced computed tomography (CT) and positron emission tomography (PET) combined with CT (PET/CT) acquired 24 hours after RF ablation. MATERIALS AND METHODS There were 45 metastases in 20 patients treated. After 24 hours, imaging of the ablation zones was performed with contrast-enhanced PET/CT. Three independent radiologists prospectively assessed contrast-enhanced CT and combined PET/CT images to identify three patterns: pattern I, no tissue enhancement or fluorodeoxyglucose uptake between the ablation zone and the liver parenchyma; pattern II, a rimlike pattern; and pattern III, a peripheral nodule. PET/CT images obtained after 8-10 weeks were evaluated for LTP. The patterns were analyzed for their sensitivity, specificity, positive predictive value, and negative predictive value for predicting LTP. RESULTS Pattern I was most frequently observed (81% for contrast-enhanced CT and 61% for PET/CT) as well as for ablation zones that showed LTP (52% and 37%, respectively). Conversely, pattern II was observed for tumors that were completely ablated (6% and 29%, respectively). Patterns II and III together had the highest sensitivity for predicting LTP (48% and 63%, respectively); pattern III had the highest specificity (94% and 95%, respectively). For nodular patterns, test characteristics were better for PET/CT compared with contrast-enhanced CT, but the difference was not significant. Nodular patterns > 1 cm achieved high positive predictive value (both 100%). CONCLUSIONS Inflammation and hyperemia can hinder interpretation on imaging 24 hours after RF ablation, especially on PET/CT. Nodular patterns around the ablation zone on early contrast-enhanced CT and PET/CT have a high predictive value for LTP and should be taken into account for disease management.

Dual-energy CT after radiofrequency ablation of liver, kidney, and lung lesions: a review of features

AbstractEarly detection of residual tumour and local tumour progression (LTP) after radiofrequency (RF) ablation is crucial in the decision whether or not to re-ablate. In general, standard contrast-enhanced computed tomography (CT) is used to evaluate the technique effectiveness; however, it is difficult to differentiate post-treatment changes from residual tumour. Dual-energy CT (DECT) is a relatively new technique that enables more specific tissue characterisation of iodine-enhanced structures because of the isolation of iodine in the imaging data. Necrotic post-ablation zones can be depicted as avascular regions by DECT on greyscale- and colour-coded iodine images. Synthesised monochromatic images from dual-energy CT with spectral analysis can be used to select the optimal keV to achieve the highest contrast-to-noise ratio between tissues. This facilitates outlining the interface between the ablation zone and surrounding tissue. Post-processing of DECT data can lead to an improved characterisation and delineation of benign post-ablation changes from LTP. Radiologists need to be familiar with typical post-ablation image interpretations when using DECT techniques. Here, we review the spectrum of changes after RF ablation of liver, kidney, and lung lesions using single-source DECT imaging, with the emphasis on the additional information obtained and pitfalls encountered with this relatively new technique. Teaching Points•Technical success of RF ablation means complete destruction of the tumour.•Assessment of residual tumour on contrast-enhanced CT is hindered by post-ablative changes.•DECT improves material differentiation and may improve focal lesion characterisation.•Iodine maps delineate the treated area from the surrounding parenchyma well.

B-LUT: Fast and low memory B-spline image interpolation

We propose a fast alternative to B-splines in image processing based on an approximate calculation using precomputed B-spline weights. During B-spline indirect transformation, these weights are efficiently retrieved in a nearest-neighbor fashion from a look-up table, greatly reducing overall computation time. Depending on the application, calculating a B-spline using a look-up table, called B-LUT, will result in an exact or approximate B-spline calculation. In case of the latter the obtained accuracy can be controlled by the user. The method is applicable to a wide range of B-spline applications and has very low memory requirements compared to other proposed accelerations. The performance of the proposed B-LUTs was compared to conventional B-splines as implemented in the popular ITK toolkit for the general case of image intensity interpolation. Experiments illustrated that highly accurate B-spline approximation can be obtained all while computation time is reduced with a factor of 5-6. The B-LUT source code, compatible with the ITK toolkit, has been made freely available to the community.

Development and validation of an automated and marker-free CT-based spatial analysis method (CTSA) for assessment of femoral hip implant migration In vitro accuracy and precision comparable to that of radiostereometric analysis (RSA)

Background and purpose — We developed a marker-free automated CT-based spatial analysis (CTSA) method to detect stem-bone migration in consecutive CT datasets and assessed the accuracy and precision in vitro. Our aim was to demonstrate that in vitro accuracy and precision of CTSA is comparable to that of radiostereometric analysis (RSA). Material and methods — Stem and bone were segmented in 2 CT datasets and both were registered pairwise. The resulting rigid transformations were compared and transferred to an anatomically sound coordinate system, taking the stem as reference. This resulted in 3 translation parameters and 3 rotation parameters describing the relative amount of stem-bone displacement, and it allowed calculation of the point of maximal stem migration. Accuracy was evaluated in 39 comparisons by imposing known stem migration on a stem-bone model. Precision was estimated in 20 comparisons based on a zero-migration model, and in 5 patients without stem loosening. Results — Limits of the 95% tolerance intervals (TIs) for accuracy did not exceed 0.28 mm for translations and 0.20° for rotations (largest standard deviation of the signed error (SDSE): 0.081 mm and 0.057°). In vitro, limits of the 95% TI for precision in a clinically relevant setting (8 comparisons) were below 0.09 mm and 0.14° (largest SDSE: 0.012 mm and 0.020°). In patients, the precision was lower, but acceptable, and dependent on CT scan resolution. Interpretation — CTSA allows detection of stem-bone migration with an accuracy and precision comparable to that of RSA. It could be valuable for evaluation of subtle stem loosening in clinical practice.

The pythagorean averages as group images in efficient groupwise registration

Many applications in medical image processing can benefit from robust and unbiased groupwise registration. However, no obvious solution is available for multimodal registration problems involving a large number of images. A technique that is frequently applied calculates the sum of the pairwise similarities between a group image and all the images in the group. This allows the algorithm to scale linearly with respect to the number of images involved. Typically the arithmetic average is used as the group image, which has been shown to be a poor choice. We present geometric and harmonic averaging as an alternative and validate their performance in mono- and multimodal experiments. These show an increased robustness and accuracy compared to the arithmetic average.