Two‐compartment modeling of tissue microcirculation revisited. Issue 5 (17th April 2017)

• Record Type: Journal Article
• Authors: Brix, Gunnar; Salehi Ravesh, Mona; Griebel, Jürgen
• Is Part Of: Medical physics. Volume 44:Issue 5(2017)
• Page Start: 1809
• Page End: 1822
• Publication Date: 2017-04-17
• Subjects: dynamic‐contrast enhanced imaging -- flow bias -- microcirculation -- two‐compartment modeling; Medical physics -- Periodicals; Medical physics; Geneeskunde; Natuurkunde; Toepassingen; Biophysics; Periodicals; Periodicals; Electronic journals; 610.153
• DOI: 10.1002/mp.12196
• Languages: English
• ISSNs: 0094-2405

Abstract

Abstract : Purpose: Conventional two‐compartment modeling of tissue microcirculation is used for tracer kinetic analysis of dynamic contrast‐enhanced (DCE) computed tomography or magnetic resonance imaging studies although it is well‐known that the underlying assumption of an instantaneous mixing of the administered contrast agent (CA) in capillaries is far from being realistic. It was thus the aim of the present study to provide theoretical and computational evidence in favor of a conceptually alternative modeling approach that makes it possible to characterize the bias inherent to compartment modeling and, moreover, to approximately correct for it. Methods: Starting from a two‐region distributed‐parameter model that accounts for spatial gradients in CA concentrations within blood‐tissue exchange units, a modified lumped two‐compartment exchange model was derived. It has the same analytical structure as the conventional two‐compartment model, but indicates that the apparent blood flow identifiable from measured DCE data is substantially overestimated, whereas the three other model parameters (i.e., the permeability‐surface area product as well as the volume fractions of the plasma and interstitial distribution space) are unbiased. Furthermore, a simple formula was derived to approximately compute a bias‐corrected flow from the estimates of the apparent flow and permeability‐surface area product obtained by model fitting. To evaluate the accuracy of the proposed modeling and bias correction method, representative noise‐free DCE curves were analyzed. They were simulated for 36 microcirculation and four input scenarios by an axially distributed reference model. Results: As analytically proven, the considered two‐compartment exchange model is structurally identifiable from tissue residue data. The apparent flow values estimated for the 144 simulated tissue/input scenarios were considerably biased. After bias‐correction, the deviations between estimated and actual parameter values were (11.2 ± 6.4) % (vs. (105 ± 21) % without correction) for the flow, (3.6 ± 6.1) % for the permeability‐surface area product, (5.8 ± 4.9) % for the vascular volume and (2.5 ± 4.1) % for the interstitial volume; with individual deviations of more than 20% being the exception and just marginal. Increasing the duration of CA administration only had a statistically significant but opposite effect on the accuracy of the estimated flow (declined) and intravascular volume (improved). Conclusions: Physiologically well‐defined tissue parameters are structurally identifiable and accurately estimable from DCE data by the conceptually modified two‐compartment model in combination with the bias correction. The accuracy of the bias‐corrected flow is nearly comparable to that of the three other (theoretically unbiased) model parameters. As compared to conventional two‐compartment modeling, this feature constitutes a major advantage for tracer kinetic analysis of both preclinical and clinical DCE imaging studies.