Imaging-based biomarker research is gaining momentum in oncology due to its non-invasive nature and the potential for monitoring tumor dynamics during treatment. The currently available biomarkers for immune checkpoint inhibitors (ICIs), such as programmed cell death ligand 1 (PD-L1) expression, tumor mutational burden, and tumor-infiltrating lymphocytes, have inherent limitations in predicting treatment outcomes. Different imaging-based biomarkers emerged as potential predictors of therapy response across various cancers. Through artificial intelligence radiomics tools, multiple quantitative imaging features are repeatedly extracted and analyzed. These non-invasive imaging features can be integrated with other predictive clinical markers to evaluate disease response.