Addressing lung truncation in 99mTc-MAA SPECT/CT for 90Y microsphere radioembolization treatment planning

Abstract

Listen

Translate article

BackgroundPrior studies have established that macroaggregated albumin (MAA)-SPECT/CT offers more robust lung shunt fraction (LSF) and lung mean absorbed dose (LMD) estimates in 90Y radioembolization in comparison to planar imaging. However, incomplete SPECT/CT coverage of the lungs is common due to clinical workflows, complicating its potential use for LSF and LMD calculations. In this work, lung truncation in MAA-SPECT/CT was addressed via correction strategies to improve 90Y treatment planning.MethodsLung truncation was simulated in 56 cases with adequate (> 90%, mean: 98%) lung coverage in MAA-SPECT/CT by removing slices in ~ 5 mm increments from the lung apices to the diaphragm. A wide range of lung coverages from 100% to < 1% in ~ 2% increments were created. LSF and LMD were calculated with four methods. (1) 2D planar imaging standard (not truncated), truncated lung SPECT/CT data was: (2) used with no corrections (SPECTTrunc), (3) uniformly extrapolated to full lung coverage (SPECTUniform), (4) fit with an empirical model to predict lung counts at full lung coverage (SPECTFit). To determine counts for LSF, full lung volumes, those modified at the lung/liver boundary (Lungs 2-cm), and those isolated to the left lung (Left Lung) were used. The correction methods were then applied to 31 independent cases without full lung coverage (< 90%, mean: 74%). The variations in LSF and LMD estimates from each correction method were analyzed.ResultsAveraged across simulated lung coverages from 40 to 80%, percent errors relative to non-truncated data for SPECTTrunc were (mean ± σ) − 22% ± 15% for LSF and 34% ± 29% for LMD. SPECTUniform had similar errors with 29% ± 26% for both LSF and LMD. SPECTFit yielded the most accurate and precise estimates for LSF and LMD, with errors of 11% ± 20% for both. The Left Lung approach equalized LMD errors in all three correction methods, with percent errors of 3% ± 17% (SPECTTrunc), 2% ± 17% (SPECTUniform), and 4% ± 13% (SPECTFit). In the 31 cases without ground truth LSF or LMD, Left Lung produced highly comparable LMD estimates, with a mean (max) coefficient of variation across the three correction methods of 4% (20%).ConclusionLSF and LMD can be estimated for 90Y radioembolization using truncated lung coverage data in MAA-SPECT/CT. Empirical models to predict lung counts at full lung coverage produced LSF and LMD estimates with minimal bias and uncertainty. With lung/liver boundary adjustments, all SPECT/CT methods assessed in this work yielded LMD estimates comparable to ground truth, even down to 50% lung coverage.