Comparison of perfusion models for quantitative T1 weighted DCE-MRI of rectal cancer

Abstract

In this work, the two compartment exchange model and two compartment uptake model were applied to obtain quantitative perfusion parameters in rectum carcinoma and the results were compared to those obtained by the deconvolution algorithm. Eighteen patients with newly diagnosed rectal carcinoma underwent 3 T MRI of the pelvis including a T1 weighted dynamic contrastenhanced (DCE) protocol before treatment. Mean values for Plasma Flow (PF), Plasma Volume (PV) and Mean Transit Time (MTT) were obtained for all three approaches and visualized in parameter cards. For the two compartment models, Akaike Information Criterion (AIC) and were calculated. Perfusion parameters determined with the compartment models show results in accordance with previous studies focusing on rectal cancer DCE-CT (PF2CX = 68 ± 44 ml/100 ml/min, PF2CU = 55 ± 36 ml/100 ml/min) with similar fit quality (AIC:169 ± 81/179 ± 77, :10 ± 12/9 ± 10). Values for PF are overestimated whereas PV and MTT are underestimated compared to results of the deconvolution algorithm. Significant differences were found among all models for perfusion parameters as well as between the AIC and values. Quantitative perfusion parameters are dependent on the chosen tracer kinetic model. According to the obtained parameters, all approaches seem capable of providing quantitative perfusion values in DCE-MRI of rectal cancer.

Quality‐based pharmacokinetic model selection on DCE‐MRI for characterizing orbital lesions

“…We tested not only the efficacy of the different models, but also their robustness and quality in order to evaluate their clinical relevance. The indicators of the quality of the modeling have been widely studied in the literature . However, there is no consensus regarding the best‐quality indicators and how to include them into a fitting strategy, and even their use remains controversial .…”

“…In our study, all six models of various complexity could distinguish benign from malignant lesions. The simplest models or even a model‐free approach had the same accuracy discriminating lesions, but the low quality of fitting of the simplest models and the high amount of information lost during modeling advocates against their use . They should probably be chosen only in simpler tissue types or when the data quality is inadequate to resolve all the features of the more complex models, as with acquisitions with low temporal resolution .…”

“…The indicators of the quality of the modeling 19,[24][25][26] have been widely studied in the literature. 9,12,17,19,24,[28][29][30][31] However, there is no consensus regarding the best-quality indicators and how to include them into a fitting strategy, 9,28 and even their use remains controversial. 31,32 Overall, some conclusions could be reached based on our study.…”

“…There remains, however, a debate on which pharmacokinetic model should be used, and most experts agree that models should be adapted to different tissues and pathologies . Model selection is thus an essential requirement in pharmacokinetic analysis of the characteristic signals . Only a few studies have evaluated DCE‐MRI in the orbit, showing its usefulness when detecting and diagnosing orbital lesions, either alone or in combination with diffusion‐weighted imaging .…”

“…9 Model selection is thus an essential requirement in pharmacokinetic analysis of the characteristic signals. [10][11][12] Only a few studies have evaluated DCE-MRI in the orbit, showing its usefulness when detecting and diagnosing orbital lesions, either alone or in combination with diffusion-weighted imaging. [12][13][14][15][16][17][18] However, these studies used an arbitrary pharmacokinetic model, without evaluating the most appropriate one for this specific application.…”