Clinical evaluation of Chang attenuation-correction parameterization for CT-free 177Lu SPECT organ quantification: implications for safety dosimetry.

Abstract

Quantitative post-therapy 177Lu SPECT depends on reliable attenuation correction. For centers without SPECT/CT, CT-free Chang attenuation correction is feasible, but its dependence on emission-derived body-contour thresholding requires validation. Twenty patients treated with 177Lu-DOTATATE underwent SPECT/CT at 4, 24, 48, and 96 h, with three consecutive SPECT acquisitions per time point. CTAC was the reference. Identical CT-defined VOIs were applied to CTAC, no attenuation correction (NC), and Chang configurations C1-C5 generated with thresholds 1-5 and μ = 0.12 cm-1. VOI-derived counts for the kidneys, liver, and total imaged body VOI were compared with CTAC using signed/absolute percentage error, Bland-Altman analysis, and error classification. NC showed systematic underestimation (mean signed error, -72.0%; mean absolute percentage error [APE], 72.1%). Chang correction improved agreement but was threshold dependent. C2 had the closest CTAC match (mean signed error, -9.2%; mean APE, 17.6%) and the highest proportions within ±10%, ±20%, and ±30% of CTAC (40.0%, 68.1%, and 79.4%). C4 shifted further toward underestimation, supporting progressive contour-shrinkage effects. Patient-level sensitivity analysis confirmed significant APE differences across C1-C5 (Friedman χ2 = 58.6, p = 5.71 × 10-12). C2 was the most balanced CTAC-matching Chang configuration in this cohort, whereas C1 may be useful for conservative kidney-focused safety assessment. These findings support local commissioning of CT-free Chang parameters but do not replace calibrated absorbed-dose calculation or fully CT-free VOI validation.