Radiomics on ¹⁷⁷Lu-DOTATATE Posttreatment Scans: Feasibility, Preprocessing Optimization, and Planar-SPECT Comparison

Area of Science:

• Nuclear Medicine
• Radiology
• Oncology

Background:

• ¹⁷⁷Lu-DOTATATE peptide receptor radionuclide therapy (PRRT) is a key treatment for neuroendocrine tumors (NETs).
• Radiomics, the extraction of quantitative features from medical images, has been primarily developed using ⁶⁸Ga-DOTATOC PET, leaving PRRT-related scintigraphy underexplored.
• This study addresses the gap by investigating radiomic feature extraction from post-PRRT ¹⁷⁷Lu-DOTATATE scans.

Purpose of the Study:

• To evaluate the feasibility of extracting hepatic radiomic features from posttreatment whole-body planar scintigraphic (WBS) and SPECT scans.
• To optimize preprocessing strategies for radiomic analysis of ¹⁷⁷Lu-DOTATATE scans.
• To compare radiomic profiles between WBS and SPECT imaging modalities.

Main Methods:

• Retrospective analysis of 183 WBS and 108 SPECT scans from 49 patients with gastroenteropancreatic NETs undergoing ¹⁷⁷Lu-DOTATATE PRRT.
• Image-intensity normalization using spleen (S-norm) and third lumbar vertebra (V-norm) mean counts.
• Radiomic feature extraction with 64-bin discretization and relative intensity rescaling, without spatial resampling.

Main Results:

• Successful texture extraction from all WBS and SPECT images.
• V-norm normalization showed more effective intensity normalization with lower variability compared to S-norm.
• Robust correlations were found for histogram-based and second-order features between S-norm and V-norm, while intensity-based features showed weak correlations.
• Intensity-based features correlated well between WBS and SPECT, but histogram-based and second-order features did not show significant correlation.

Conclusions:

• Radiomic analysis of posttreatment ¹⁷⁷Lu-DOTATATE scans is feasible.
• Third lumbar vertebra normalization (V-norm) is the preferred method for intensity normalization due to reduced variability.
• Radiomic features from WBS and SPECT are not directly comparable, except for intensity-based features.