Rgp Richard Lopata

On the identifiability of pharmacokinetic parameters in dynamic contrast‐enhanced imaging

The so‐called “Kety model” is a two‐compartment pharmacokinetic model describing tumor perfusion kinetics. Its parameters, the transendothelial transfer constant (Ktrans), extravascular extracellular volume fraction (υe), and microvascular plasma volume fraction (υp), can be estimated with dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI). However, the results obtained by current methods show large variation in predictability and reliability. Here, the aim was to examine which experimental conditions have to be fulfilled to avoid large uncertainties and mutual dependencies of the parameters. Using frequency response analysis and simulation, the identifiability of the model was examined. The requirements and influence of contrast enhancement measurements, such as temporal resolution, signal to noise ratio, and contrast injection rate, on the accuracy of the parameters were analyzed. Tissue response characteristics revealed a low‐frequency system with a cutoff frequency equal to Ktrans/υe, which confines the required temporal resolution. For malignant tissue with hyperpermeable vasculature (high Ktrans) a higher sampling frequency is required to accurately estimate Ktrans than for normal tissue. Too low sampling rates or too low injection rates resulted in inaccurate Ktrans values and hereby unreliable classification of malignant tissue. Magn Reson Med 58:425–429, 2007.

Predicting Target Displacements Using Ultrasound Elastography and Finite Element Modeling

Soft tissue displacements during minimally invasive surgical procedures may cause target motion and subsequent misplacement of the surgical tool. A technique is presented to predict target displacements using a combination of ultrasound elastography and finite element (FE) modeling. A cubic gelatin/agar phantom with stiff targets was manufactured to obtain pre- and post-loading ultrasound radio frequency (RF) data from a linear array transducer. The RF data were used to compute displacement and strain images, from which the distribution of elasticity was reconstructed using an inverse FE-based approach. The FE model was subsequently used to predict target displacements upon application of different boundary and loading conditions to the phantom. The influence of geometry was investigated by application of the technique to a breast-shaped phantom. The distribution of elasticity in the phantoms as determined from the strain distribution agreed well with results from mechanical testing. Upon application of different boundary and loading conditions to the cubic phantom, the FE model-predicted target motion were consistent with ultrasound measurements. The FE-based approach could also accurately predict the displacement of the target upon compression and indentation of the breast-shaped phantom. This study provides experimental evidence that organ geometry and boundary conditions surrounding the organ are important factors influencing target motion. In future work, the technique presented in this paper could be used for preoperative planning of minimally invasive surgical interventions.

ECHO-COMPUTED TOMOGRAPHY STRAIN IMAGING OF HEALTHY AND DISEASED CAROTID SPECIMENS

To improve our understanding of the mechanical behavior of human atherosclerotic plaque tissue, fully 3-D geometrical, morphological and dynamical information is essential. For this purpose, four-dimensional (3-D+t) strain imaging using an ultrasound tomography approach (echo-computed tomography) was performed in carotid arteries in vitro. The method was applied to a carotid phantom (CPh), a porcine carotid artery (PC) and human carotid atherosclerotic plaque samples (HC, n = 5). Each sample was subjected to an intraluminal pressure, after which 2-D longitudinal ultrasound images were obtained for 36 angles along the circumferential direction. Local deformations were estimated using a 2-D strain algorithm, and 3-D radial strain data were reconstructed. At systole, median luminal strains of 15% (CPh) and 18% (PC) were found, which is in agreement with the stiffness of the material and applied pressure pulse. The elastographic signal-to-noise ratio was consistent in all directions and ranged from 16 to 36 dB. Furthermore, realistic but more complex strain patterns were found for the HC, with 99th percentile systolic strain values ranging from 0.1% to 18%.

VASCULAR ELASTOGRAPHY: A VALIDATION STUDY

Vascular elastography techniques are promising tools for mechanical characterization of diseased arteries. These techniques are usually validated with simulations or phantoms or by comparing results with histology or other imaging modalities. In the study described here, vascular elastography was applied to porcine aortas in vitro during inflation testing (n = 10) and results were compared with those of standard bi-axial tensile testing, a technique that directly measures the force applied to the tissue. A neo-Hookean model was fit to the stress-strain data, valid for large deformations. Results indicated good correspondence between the two techniques, with GUS = 110 ± 11 kPa and GTT = 108 ± 10 kPa for ultrasound and tensile testing, respectively. Bland-Altman analysis revealed little bias (GUS-GTT = 2 ± 20 kPa). The next step will be the application of a non-linear material model that is also adaptable for in vivo measurements.

Towards mechanical characterization of intact endarterectomy samples of carotid arteries during inflation using Echo-CT

In this study, an experimental framework is described that allows pressurization of intact, human atherosclerotic carotid samples (inflation testing), in combination with ultrasound imaging. Eight fresh human carotid endarterectomy samples were successfully pressurized and tested. About 36 2-D (+t) ultrasound datasets were acquired by rotating the vessel in 10° steps (Echo-CT), from which both 3-D geometry and 3-D strain data were obtained. Both geometry and morphology were assessed with micro-CT imaging, identifying calcified and lipid rich regions. US-based and CT-based geometries were matched for comparison and were found to show good agreement, with an average similarity index of 0.71. Realistic pressure-volume relations were found for 6 out of 9 samples. 3-D strain datasets were reconstructed, revealing realistic strain patterns and magnitudes, although the data did suffer from a relatively high variability. The percentage of fat and calcifications (micro-CT) were compared with the median, 75th and 99th percentile strain values (Echo-CT). A moderate trend was observed for 75th and 99th percentile strains, higher strains were found for more lipid rich plaques, where lower strains were found for highly calcified plaques. However, an inverse numerical modeling technique is necessary for proper mechanical characterization the of plaque components, using the geometry, morphology and wall deformation as input.

Three-dimensional ultrasound strain imaging of skeletal muscles

Muscle contraction is characterized by large deformation and translation, which requires a multi-dimensional imaging modality to reveal its behavior. Previous work on ultrasound strain imaging of the muscle contraction was limited to 2D and bi-plane techniques. In this study, a three-dimensional (3D) ultrasound strain imaging technique was tested against 2D strain imaging and used for quantifying deformation of skeletal muscles. A phantom compression study was conducted for an experimental validation of both 2D and 3D methods. The phantom was compressed 3% vertically and pre- and post-compression full volume radio frequency (RF) ultrasound data were acquired using a matrix array transducer. A cross-correlation-based algorithm using either 2D or 3D kernels was applied to obtain the displacement estimates. These estimates were converted to Cartesian space and subsequently, strain was derived using a least-squares strain estimator (LSQSE). The 3D results were compared with the 2D results and the theoretically predicted displacement and strain. Comparison between 2D and 3D kernels was performed on data from a plane with a large tilt angle to study the influence of out-of-plane motion on the two techniques. To demonstrate the in vivo feasibility, 3D strain was calculated from live 3D data, acquired during a 2 second isometric contraction and relaxation of the quadriceps muscle in a healthy volunteer. The phantom study showed good correlation between estimated displacements and the theoretically predicted displacements. Root-mean-squared errors (RMSE) were 0.16, 0.17 and 0.13 mm in the x-, y- and z-direction respectively. The absolute RMSE for the 3D strain values were 0.94, 1.2 and 0.41% in the x-, y- and z-direction respectively. The 2D method performed worse, with 3 (x-direction) to 6 (z-direction) times higher RMSE values. The larger errors in lateral and elevational direction with respect to the axial RMSE are potentially caused by the large angle between the ultrasound beams. Initial in vivo results revealed 3D strain curves which clearly visualized the contraction and relaxation of the quadriceps muscles. Muscle deformation estimation using real-time 3D ultrasound RF-data seems feasible and the use of 3D kernels improves displacement estimation in comparison to 2D techniques. Future work will focus on improving lateral and elevational displacement estimation, and investigating local differences of strain in skeletal muscles and its clinical relevance.

Visualization of vasculature using a hand-held photoacoustic probe: phantom and in vivo validation

Abstract. Assessment of microvasculature and tissue perfusion can provide diagnostic information on local or systemic diseases. Photoacoustic (PA) imaging has strong clinical potential because of its sensitivity to hemoglobin. We used a hand-held PA probe with integrated diode lasers and examined its feasibility and validity in the detection of increasing blood volume and (sub) dermal vascularization. Blood volume detection was tested in custom-made perfusion phantoms. Results showed that an increase of blood volume in a physiological range of 1.3% to 5.4% could be detected. The results were validated with power Doppler sonography. Using a motorized scanning setup, areas of the skin were imaged at relatively short scanning times (<10  s/cm2) with PA. Three-dimensional visualization of these structures was achieved by combining the consecutively acquired cross-sectional images. Images revealed the epidermis and submillimeter vasculature up to depth of 5 mm. The geometries of imaged vasculature were validated with segmentation of the vasculature in high-frequency ultrasound imaging. This study proves the feasibility of PA imaging in its current implementation for the detection of perfusion-related parameters in skin and subdermal tissue and underlines its potential as a diagnostic tool in vascular or dermal pathologies.

Identifiability analysis of the standard pharmacokinetic models in DCE MR imaging of tumours

The usage of dynamic contrast-enhanced MRI (DCE-MRI) as a clinical tool is still widely assessed. Application of the standard pharmacokinetic models to obtain physiologically relevant parameter values using DCE-MRI in tumours is not trivial, when the temporal resolution is low. Mathematical analysis and analysis by simulation of the identifiability for the generalized and extended Kety models was executed. Parameter estimation was executed using synthetic data sets and maximum likelihood estimation (MLE). The influence of temporal resolution was examined. The generalized and extended Kety model showed a large bias in the parameter estimates (10-120%) for sampling times >4 s, although the estimated variance was relatively low (<1%). This was in accordance with the generated contour plots of the hyperplane of the MLE cost-function. The influence of measurement noise on the input and output turned out to be less significant than the temporal resolution.

Semi-3D strain imaging in normal and LVAD supported ex vivo beating hearts

Cardiac strain imaging (CSI) is a powerful technique which is able to quantify and localize cardiac defects. However, validation and reproducibility remain major issues. In this study, semi-3D CSI was performed on ex vivo beating porcine hearts to quantify the reproducibility of CSI and examine the feasibility of clinical application of CSI for the assessment of cardiac function in patients with a left ventricular assist device. Results reveal highly reproducible strain results between heartbeats. However, differences were found between hearts, due to biological variation and condition of the hearts. Supported hearts showed a decrease in strain and strain rate, and strain pattern with higher pump speeds. Future work should include translation of the measured parameters into quantitive assessment of cardiac function and validation in a long-term patient study.

A longitudinal feasibility study on AAA growth vs. ultrasound elastography

To improve risk stratification of abdominal aortic aneurysms, ultrasound elastography could provide insight in changes in mechanical properties of the aneurysmal wall. In this feasibility study, 2-D US elastography was performed in AAA patients in a longitudinal study. Several patients with non-growing, slow growing and fast growing AAAs were imaged using 2D ultrasound. The RF-data were processed to estimate the mechanical properties, e.g., compliance (C), distensibility (D) and incremental Youngs modulus (Einc). An improved RF-based tracking algorithm was used to estimate the volume changes over time, assuming rotational symmetry of the aneurysm. The estimated parameters and possible changes were compared between the three groups. Incremental moduli ranging from 1- 8 MPa were found. The distensibility ranged from 2 to 10 MPa-1. The relative increase in Einc was -10 to 100% for the patients with no growth, 25% -125% for the medium group and 75% to 700% for the patients with fast growth. Distensibility also decreased by a factor 5. The largest increase in stiffness corresponded to the largest increase in diameter. These preliminary findings imply that large changes in mechanical properties occur in the aneurysmal wall in periods of considerable growth. Automated segmentation and higher frame rates might decrease the intra-subject variability. Finally, the inclusion of more patients is required to strengthen the evidence found.