Article

Functional CT imaging of prostate cancer

University of Toronto, Toronto, Ontario, Canada
Physics in Medicine and Biology (Impact Factor: 2.92). 10/2003; 48(18):3085-100. DOI: 10.1016/S0360-3016(01)02190-3
Source: PubMed

ABSTRACT The purpose of this paper is to investigate the distribution of blood flow (F), mean capillary transit time (Tc), capillary permeability (PS) and blood volume (vb) in prostate cancer using contrast-enhanced CT. Nine stage T2-T3 prostate cancer patients were enrolled in the study. Following bolus injection of a contrast agent, a time series of CT images of the prostate was acquired. Functional maps showing the distribution of F, Tc, PS and vb within the prostate were generated using a distributed parameter tracer kinetic model, the adiabatic approximation to the tissue homogeneity model. The precision of the maps was assessed using covariance matrix analysis. Finally, maps were compared to the findings of standard clinical investigations. Eight of the functional maps demonstrated regions of increased F, PS and vb, the locations of which were consistent with the results of standard clinical investigations. However, model parameters other than F could only be measured precisely within regions of high F. In conclusion functional CT images of cancer-containing prostate glands demonstrate regions of elevated F, PS and Vb. However, caution should be used when applying a complex tracer kinetic model to the study of prostate cancer since not all parameters can be measured precisely in all areas.

0 Followers
 · 
56 Views
  • Source
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Modeling of the perfusion using dynamic contrast-enhanced (DCE) imaging based on convolution models is gaining increasing attention. The time activity curve of ith voxel is sampled at predefined time intervals, yielding a contrast concentration Ci(t) evaluated at discrete times t=t1…tNT. Perfusion model is then specified to express Ci(t) as the convolution of an impulse response function IRF(t) and the arterial input function A(t). For a given model, IRF(t) is represented as a piecewise analytical function that is parametrized by tissue parameters such as flow rate, transit time, or distribution volume. These parameters {p1, p2, … pn} are then used to fit the measured data Ci(t) by minimizing the residual: Typically the convolution operator is invoked on the order of 104 times for each voxel’s minimization, and the imaging volume of interest contains between 105 and 106 voxels. Thus, an efficient implementation of the convolution represents a significant computational challenge. The naive approach consists of a uniform sampling t=t1, … tN of the temporal domain, extrapolation of A(t) and IRF(t) over these samples, and calculation of convolution either by discrete integration or by multiplying corresponding Fourier transforms. Computational complexity is O(N2) and O(N log(N)) respectively. In practice direct integration is often used due to its simplicity. Additionally, due to discrete sampling of IRF result in an error that is proportional to 1/N. To improve the computational speed and be able to control the convolution error due to discrete sampling of IRF(t) we have developed and implemented the adaptive convolution algorithm. The method was tested for its speed and convolution accuracy on renal perfusion renography data using Gd-DTPA as the tracer.
    ISMRM 2008, Toronoto, Canada; 05/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: Utilizing CT angiography enhances image quality in PCT, thereby permitting acquisition at ultra-low dose. Dynamic CT acquisitions were obtained at 80 kVp with decreasing tube current-time product [milliamperes x seconds (mAs)] in patients suspected of ischemic stroke, with concurrent CTA of the cervical and intracranial arteries. By utilizing fast Fourier transformation, high spatial frequencies of CTA were combined with low spatial frequencies of PCT to create a virtual PCT dataset. The real and virtual PCT datasets with decreasing mAs were compared by assessing contrast-to-noise ratio (CNR), signal-to-noise ratio (SNR), and noise and PCT values and by visual inspection of PCT parametric maps. Virtual PCT attained CNR and SNR three- to sevenfold superior to real PCT and noise reduction by a factor of 4-6 (p < 0.05). At 20 mAs, virtual PCT achieved diagnostic parametric maps, while the quality of real PCT maps was inadequate. At 10 mAs, both real and virtual PCT maps were nondiagnostic. Virtual PCT (but not real PCT) maps regained diagnostic quality at 10 mAs by applying 40 % adaptive statistical iterative reconstruction (ASIR) and improved further with 80 % ASIR. Our new method of creating virtual PCT by combining ultra-low-dose PCT with CTA information yields diagnostic perfusion parametric maps from PCT acquired at 20 or 10 mAs with 80 % ASIR. Effective dose is approximately 0.20 mSv, equivalent to two chest radiographs.
    Neuroradiology 08/2014; 56(11). DOI:10.1007/s00234-014-1416-1 · 2.37 Impact Factor