CT-Based Radiomics Predicts the Functional State of Tumor-Infiltrating CD8 T Cells and Prognosis in NSCLC.
[RATIONALE AND OBJECTIVES] The functional status of CD8 T cells is a key factor influencing the prognosis in patients with non-small cell lung cancer (NSCLC).
- p-value P < 0.001
APA
Sang B, Zang X, et al. (2026). CT-Based Radiomics Predicts the Functional State of Tumor-Infiltrating CD8 T Cells and Prognosis in NSCLC.. Academic radiology. https://doi.org/10.1016/j.acra.2026.01.057
MLA
Sang B, et al.. "CT-Based Radiomics Predicts the Functional State of Tumor-Infiltrating CD8 T Cells and Prognosis in NSCLC.." Academic radiology, 2026.
PMID
41723041
Abstract
[RATIONALE AND OBJECTIVES] The functional status of CD8 T cells is a key factor influencing the prognosis in patients with non-small cell lung cancer (NSCLC). We aimed to develop a radiomics model predicting the functional state of tumor-infiltrating CD8 T cells in NSCLC, explore semantic characteristics linking radiomic features to CD8 T cell exhaustion, and establish a prognostic nomogram.
[MATERIALS AND METHODS] A retrospective cohort of 256 patients with NSCLC undergoing radical resection with CD8 T cell functional status determined by multiplex immunofluorescence staining was randomly divided 7:3 into training and validation sets. Radiomic features from preoperative contrast-enhanced CT scans were used to develop predictive models for high density of tumor center pre-dysfunctional CD8 T cells (high-T) and high density of invasive margin dysfunctional CD8 T cells (high-T) through least absolute shrinkage and selection operator, followed by semantic analysis. A nomogram for predicting recurrence-free survival integrated radiomics models with clinical characteristics.
[RESULTS] Only the high-T radiomics model was successfully established, yielding areas under the curve of 0.933 (training) and 0.792 (validation). Peritumoral imaging features on contrast-enhanced CT (fibrosis, inflammation, and atelectasis) were associated with CD8 T cell exhaustion, evidenced by significantly higher high-T proportions: 31.6% vs. 7.5% (P < 0.001), 35.6% vs. 8.1% (P < 0.001), and 40.0% vs. 11.8% (P = 0.028). The nomogram incorporating high-T radiomics score, T stage, and N stage predicted 1- to 4-year predicting recurrence-free survival with areas under the curve of 0.733, 0.713, 0.637, and 0.600 (training), and 0.629, 0.669, 0.550, and 0.593 (validation).
[CONCLUSION] Radiomics can predict the functional exhaustion of tumor-infiltrating CD8 T cells in NSCLC, with specific imaging features associated with this process. Combining the radiomics model with clinical characteristics facilitates the assessment of patient prognosis.
[MATERIALS AND METHODS] A retrospective cohort of 256 patients with NSCLC undergoing radical resection with CD8 T cell functional status determined by multiplex immunofluorescence staining was randomly divided 7:3 into training and validation sets. Radiomic features from preoperative contrast-enhanced CT scans were used to develop predictive models for high density of tumor center pre-dysfunctional CD8 T cells (high-T) and high density of invasive margin dysfunctional CD8 T cells (high-T) through least absolute shrinkage and selection operator, followed by semantic analysis. A nomogram for predicting recurrence-free survival integrated radiomics models with clinical characteristics.
[RESULTS] Only the high-T radiomics model was successfully established, yielding areas under the curve of 0.933 (training) and 0.792 (validation). Peritumoral imaging features on contrast-enhanced CT (fibrosis, inflammation, and atelectasis) were associated with CD8 T cell exhaustion, evidenced by significantly higher high-T proportions: 31.6% vs. 7.5% (P < 0.001), 35.6% vs. 8.1% (P < 0.001), and 40.0% vs. 11.8% (P = 0.028). The nomogram incorporating high-T radiomics score, T stage, and N stage predicted 1- to 4-year predicting recurrence-free survival with areas under the curve of 0.733, 0.713, 0.637, and 0.600 (training), and 0.629, 0.669, 0.550, and 0.593 (validation).
[CONCLUSION] Radiomics can predict the functional exhaustion of tumor-infiltrating CD8 T cells in NSCLC, with specific imaging features associated with this process. Combining the radiomics model with clinical characteristics facilitates the assessment of patient prognosis.