Artificial intelligence-enabled opportunistic identification of immune checkpoint inhibitor-related adverse events using [18F]FDG PET/CT

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Purpose: Immune checkpoint inhibitors (ICI) have transformed cancer therapy, improving outcomes in malignancies like lung cancer, melanoma, and lymphoma by targeting PD-1, PD-L1, and CTLA-4 to enhance T cell-mediated tumor destruction. However, ICI often induce immune-related adverse events (irAEs) across multiple organs. [ 18 F]FDG PET/CT is a valuable tool for assessing immune activation, but whole-organ inflammation evaluation remains time-consuming and prone to variability. This study investigates AI-driven organ-level [ 18 F]FDG PET/CT uptake changes pre- and post-ICI therapy to opportunistically detect irAEs. Methods: A total of 64 patients with lung cancer, melanoma, or lymphoma who underwent [ 18 F]FDG PET/CT before and after ICI therapy were included. Automated delineation of anatomical structures was performed using an artificial intelligence approach involving multiple subsequently applied deep learning models. SUV max and SUV mean were quantified for irAE-related organs (thyroid, myocardium, pancreas, liver, bone, spleen, and adrenal glands) after automated tumor lesion removal. Statistical analysis identified uptake changes linked to clinically observed inflammation-related adverse events. Results: Among 64 patients (mean scan interval 4.5 months), thyroid uptake increased post-ICI (ΔSUV max =0.44, p = 0.04; ΔSUV mean =0.22, p = 0.08). Increased uptake occurred in 55% (SUV max ) and 53% (SUV mean ) of patients. Thyroid-related adverse events were more frequent in those with increased uptake (29% vs. 3%, p = 0.008), with uptake increase being the most pronounced in the cases of hypothyroidism ( p = 0.002 for SUV max , p = 0.03 for SUV mean ). Myocardial, pancreatic, and hepatic uptake increased but without statistical significance ( p > 0.05). Conclusion: AI-enabled opportunistic [ 18 F]FDG PET/CT screening effectively detects thyroid-related irAEs, particularly hypothyroidism, through uptake quantification. This approach offers a promising biomarker for early irAE detection, enhancing patient management and immunotherapy optimization.
Funding Information
  • Medizinische Universität Wien

This publication has 28 references indexed in Scilit: