Artificial intelligence in immunotherapy PET/SPECT imaging.

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Immunotherapy has transformed cancer treatment, but predicting patient response and side effects remains challenging.
• Traditional imaging methods may not fully capture response patterns in immunotherapy-treated patients.
• There is a need for advanced tools to evaluate treatment efficacy and predict outcomes.

Purpose of the Study:

• To investigate the role of artificial intelligence (AI) in Positron Emission Tomography (PET) and Single-Photon Emission Computed Tomography (SPECT) imaging for patients undergoing immunotherapy.
• To review how AI models are developed using medical imaging for predictive applications in cancer immunotherapy.

Main Methods:

• A scoping review of MEDLINE, CENTRAL, and Embase databases was conducted.
• Search terms included immunotherapy, PET/SPECT imaging, and AI/radiomics.
• Data were extracted from relevant articles published up to October 12, 2022.

Main Results:

• Twenty-four relevant studies were selected from 217 identified articles.
• The primary tumor types studied were lung, lymphoma, and melanoma.
• AI models, utilizing radiomics or deep learning on PET/SPECT images, were primarily used for prognostication and tumor phenotyping.
• Most studies (75%) included AI model validation.

Conclusions:

• AI-guided approaches using PET/SPECT imaging show potential for optimizing immunotherapy management.
• Further validation in large, prospective, multicenter cohorts is necessary before clinical implementation.
• AI models built from baseline imaging can predict prognosis and tumor phenotype in immunotherapy-treated patients.