Automatic Intracranial Segmentation: Is the Clinician Still Needed?

INTRODUCTION: Stereotactic hypofractionated radiotherapy is an effective treatment for brain metastases in oligometastatic patients. Its planning is however time-consuming because of the number of organs at risk to be manually segmented. This study evaluates 2 automated segmentation commercial software. METHODS: Patients were scanned in the treatment position. The computed tomography scan was registered on a magnetic resonance imaging and volumes were manually segmented by a clinician. Then 2 automated segmentations were performed (with iPlan and Smart Segmentation). RT STRUCT files were compared with Aquilab's Artistruct segment comparison module. We selected common segmented volume ratio as the main judging criterion. Secondary criteria were Dice-Sørensen coefficients, overlap ratio, and additional segmented volume. RESULTS: Twenty consecutive patients were included. Agreement between manual and automated contouring was poor. Common segmented volumes ranged from 7.71% to 82.54%, Dice-Sørensen coefficient ranged from 0.0745 to 0.8398, overlap ratio ranged from 0.0414 to 0.7275, and additional segmented volume ranged from 9.80% to 92.25%. Each software outperformed the other on some organs while performing worse on others. CONCLUSION: No software seemed clearly better than the other. Common segmented volumes were much too low for routine use in stereotactic hypofractionated brain radiotherapy. Manual editing is still needed.

T2*-relaxivity contrast imaging: first results.

In this study, T2*- relaxivity contrast imaging (RCI) is proposed for new contrast generation in MRI. The method produces images of relaxivities r*2,vasc and r*2,EES caused by susceptibility gradients across the vessel walls and cell membranes, respectively. The sensitivity to noise was assessed with a simulation study, and initial results are presented for five colorectal tumor xenografts in nude mice. Simulations show that the new relaxivity parameters are at least as accurate and precise as standard parameters such as plasma volume and interstitial volume. Mean values of both relaxivities were significantly different (r*2,vasc=10.9±2.9 mM(-1) s(-1) and r*2,EES=15.6±2.6 mM(-1) s(-1)). r*2,vasc (r=0.67) and r*2,EES (r=0.52) were weakly correlated with plasma volume and interstitial volume, respectively. Images of r*2,vasc and r*2,EES reveal a different tumor structure than plasma volume and interstitial volume maps. These results suggest that relaxivity contrast imaging is practically feasible and might offer supplementary information compared to dynamic contrast-enhanced-MRI.

Radial multigradient-echo DCE-MRI for 3D K(trans) mapping with individual arterial input function measurement in mouse tumor models.

The purpose of this study was to provide proof of concept for a new three-dimensional (3D) radial dynamic contrast enhanced MRI acquisition technique, called "Radial Entire Tumor with Individual Arterial input function dynamic contrast-enhanced MRI" (RETIA dynamic contrast-enhanced MRI), which allows for the simultaneous measurement of an arterial input function in the mouse heart at 2 s temporal resolution and coverage of the whole tumor. Alternating 2D and 3D projections contribute to the 2D heart image or 3D tumor data with a 3-cm field of view. Sixty-four 2D images of the heart are obtained during acquisition of each 3D tumor dataset. In a pilot study, global K(trans) and ve values were measured in four mice, in a respiratory motion-animated subcutaneously implanted breast tumor model. This technique is expected to be most useful for the characterization of microvasculature in motion-animated orthotopic tumors.

2D and 3D radial multi-gradient-echo DCE MRI in murine tumor models with dynamic R*2-corrected R1 mapping.

Dynamic contrast-enhanced MRI is extensively studied to define and evaluate biomarkers for early assessment of vasculature-targeting therapies. In this study, two-dimensional and three-dimensional radial multi-gradient-echo techniques for dynamic R*(2)-corrected R(1) mapping based on the spoiled gradient recalled signal equation were implemented and validated at 4.7 T. The techniques were evaluated on phantoms and on a respiratory motion animated tumor model. R(1) measurements were validated with respect to a standard inversion-recovery spin-echo sequence in a four-compartment phantom covering a range of relaxation rates typically found in tumor tissue. In the range of [0.4, 3] sec(-1), R(1) differences were less than 10% for both two-dimensional and three-dimensional experiments. A dynamic contrast-enhanced MRI pilot study was performed on a colorectal tumor model subcutaneously implanted in mice at the abdominal level. Low motion sensitivity of radial acquisition allowed image recording without respiratory triggering. Three-dimensional K(trans) maps and significantly different mean K(trans) values were obtained for two contrast agents with different molecular weights. The radial multi-gradient-echo approach should be most useful for preclinical experimental conditions where the tissue of interest experiences physiologic motion, like spontaneous extracerebral tumors developed by transgenic mice, and where dynamic contrast-enhanced MRI is performed with high-relaxivity contrast agents.

In vitro setup to study permeability characteristics of contrast agents by MRI.

An experimental setup consisting of a hollow fiber module (HFM) was developed for the in vitro study of contrast agent (CA) permeability. Controllable flow and known fiber characteristics allowed permeability studies under well-defined conditions with CAs of different molecular weight (MW). In the MRI experiments performed at 4.7 T, the system was perfused at a constant flow rate (5 ml/min) with water and four CA of different MW: Gd-DOTA (MW=0.6 kDa), P846 (3.5 kDa), P792 (6.5 kDa) and P717 (50.5 kDa). R(1) time courses were measured with a saturation-recovery multi-gradient-echo snapshot sequence in the fiber-free HFM input and the fiber-filled center. Concentration time courses were calculated, and CA extravasation was analyzed with a pharmacokinetic model yielding exchange rate constant k(ie). Only Gd-DOTA (k(ie)=2.37+/-0.16 min(-1)) and P846 (k(ie)=0.58+/-0.17 min(-1)) showed quantifiable extravasation. P717 perfusion yielded an intra-capillary volume fraction of 15.6+/-2.7% compared with 12% estimated from the HFM manufacturer's specifications. In conclusion, the experimental setup allowed classification of in vitro permeability characteristics for CAs with different MW and therefore holds potential for systematic comparison of CAs currently under development.

Simultaneous dynamic T1 and T2* measurement for AIF assessment combined with DCE MRI in a mouse tumor model.

A double-delay SR-MGE-SNAP sequence allowing simultaneous T1 and T2* measurement was developed for integrating arterial input function (AIF) measurement into DCE MRI. Implemented on a 4.7-T animal MR system, this technique was applied to mice with colorectal tumor xenografts. AIF, measured in the mouse heart, was modeled by a bi-exponential function, whereas tumor K(trans) and v(e) parameter maps were obtained from analysis with a two- compartment model using an individually measured AIF. AIF analysis of T2*-corrected data yielded A1 = 9.2 +/- 4.3 kg/l, A(2) = 4.2 +/- 0.8 kg/l, m1 = 2.3 +/- 1.1 min(-1), and m2 = 0.05 +/- 0.02 min(-1). The mean initial plasma concentration C ( p )(t = 0) = 8.0 +/- 2.7 mM was compatible with estimated 8.6 mM. Without T2*-correction distribution phase parameters A1, m1, and C(p)(t = 0) were underestimated. In tumors, neglect of T2* effects yielded mean K(trans) values which were reduced by 14% (P < 0.05), whereas v(e) showed only a slight non-significant reduction. Simultaneous measurement of DeltaR1 and DeltaR2* studied in highly and poorly vascularized and (pre-)necrotic tumor regions revealed complementary behavior of both parameters with respect to vascular properties. In conclusion, the presented measurement technique is a promising tool for dynamic MRI applications studied in animal models at high field strengths and/or with CA of high relaxivities, as it combines classical DCE MRI integrating AIF assessment with dynamic T2* measurement.