Proof of Concept Clinical Evaluation of Gradient Different Weight Threshold Segmentation for Accurate Tumour Delineation in I-131 SPECT Imaging

Objectives: We developed a modified gradient-different-weight threshold method, adapted from gradient-weighted thresholding, to improve robustness over fixed-percentage approaches. Material and Methods: The method computes voxel-wise gradient weights and optimises a threshold value by calibrating against a NEMA IQ phantom (six spheres) across sphere-to-background ratios (12:1, 10:1, 8:1, 6:1), linking the optimised threshold to a ratio factor (Cmax/Cmean) while comparing three background-estimation strategies (multi-circle, ULWL-guided narrow annular, and FWHM-based). Phantom optimisation showed the ULWL-guided narrow annular strategy produced the most consistent threshold-versus-ratio relationship. A proof-of-concept clinical evaluation on two retrospective I-131 SPECT/CT datasets used reference volumes from two nuclear medicine physicians (average contour) and reported percentage error, RMSD, and Dice similarity coefficient (DSC). Results: For a thyroid-only case, the method yielded 27.1 mL versus 25.3 mL (9.71% error; DSC = 0.9541). In a whole-body multi-lesion case, delineation became less reliable when uptake was diffuse, lesions were low-count, or lesions were closely spaced, which elevated local background estimates. Conclusion: These results suggest gradient-different-weight thresholding can improve I-131 SPECT tumour segmentation when background selection is feasible. However, larger multi-patient studies and direct side-by-side benchmarking are required to define its clinical operating range.