Fast quantitative determination of lung shunt fraction using orthogonal planar projections in hepatic radioembolization

Objectives Hepatic radioembolization (RE) procedures are preceded by a 99mTc-MAA scout dose, which is used to predict extra-hepatic shunting. Currently, the lung shunt fraction (LSF) is estimated using the geometric mean of conjugate view planar scintigraphy (PS-GM). However, concern has been raised about possible LSF overestimation using PS-GM. This potentially leads to patients unnecessarily being excluded from subsequent RE treatment. Conversely, SPECT/CT is considered to be the gold standard but is also lengthy, since two SPECT/CTs are required to cover both the liver and the lungs. In addition, the requirement of labor-intensive segmentation procedures, make SPECT/CT a suboptimal method for fast LSF estimation in clinical practice. The objective of this study is to develop a practical method that combines the quantitative qualities of SPECT/CT with the speed and efficiency of planar scintigraphy for accurate LSF estimation.

Methods To accurately correct for the lack of attenuation, scatter and collimator/detector effects, ideally the activity in three dimensions has to be estimated. To achieve this in planar scintigraphy, information from the sagittal plane was added to the coronal plane information from PS-GM by an additional orthogonal set of conjugate views. Based on these whole body projections, a coarse 3D volume was reconstructed incorporating all relevant physics using a Monte Carlo based reconstruction algorithm. This method is called Quantitative Orthogonal Conjugate Planar method (QOCP). For validation, the 4D XCAT digital anthropomorphic phantom was used to Monte Carlo simulate SPECT/CT and planar scintigraphy data sets. Both breathing and static situations were evaluated for a range of LSFs. Noise levels were chosen to match total clinical acquisition times of 10 minutes (70 cm PS-GM), two times 10 min (70 cm QOCP) and 50 min (2 SPECT/CT acquisitions). In all cases 150 MBq in the FOV was simulated. For analysis of QOCP, 3D VOIs were determined by backprojection of 2D delineations on the two views (sagittal and coronal). This resulted in VOIs in the same 3D space as the activity reconstruction. SPECT/CT was analyzed by delineating a 3D VOI slice-by-slice. For PS-GM, 2D liver and lung ROIs were delineated on the planar data. LSF was determined from these regions.

Results The simulations showed that PS-GM overestimated LSF up to 47% relative to ground truth. Breathing motion slightly lessened the maximum overestimation to 40%. SPECT/CT-based LSF estimations were more accurate, with at most 8% overestimation, and up to 5% underestimated activity in the breathing case. Our newly proposed QOCP method resulted in errors up to 2% underestimation in the static situation and up to 6% underestimation in the breathing case, for the clinical range of LSFs.

Conclusions PS-GM LSF overestimation is large. It is therefore suggested to add a second conjugate view measurement, without requiring lengthy SPECT/CTs. This QOCP method adds enough volumetric information to allow for attenuation correction and spatial specificity and is simple enough to require only two (manual) 2D delineations. This enables fast and quantitatively accurate LSF estimation in a clinical setting. $$graphic_5078EAD0-FDFA-4DAE-B2CE-1C2CDB3BD390$$