2023 North America Conference on Lung Cancer (NACLC) - Posters and Abstracts-PP01.044 Fadila Zerka NACLC23 Abstract

PP01.044 Fadila Zerka NACLC23 Abstract

Pdf Summary

Researchers conducted a study to explore the use of CT imaging biomarkers to stratify patients based on CD8 lymphocyte infiltration levels in non-small cell lung cancer (NSCLC). CD8 T cell infiltration has been shown to correlate with improved outcomes and overall survival in immunotherapy, but current methods rely on invasive biopsies. The study utilized four publicly available datasets of NSCLC patients with CT imaging and RNA-Seq data. Lung lesions were segmented using a semi-automatic tool, and radiomics features were extracted. Features with near zero variance, high correlation, and linear combinations were eliminated from analysis. Three data combinations were tested for reproducibility. A logistic regression model with Elastic Net regularization was used to classify CD8 cell infiltration levels. <br /><br />The results showed that different subsets of features were able to differentiate CD8 groups in the three use-cases. Four texture features consistently discriminated between CD8 groups. The model trained on these discriminative features achieved a mean area under the curve (AUC-ROC) of 0.73 with a standard deviation of 0.08, and an AUC-ROC of 0.67 (with a 95% confidence interval of 53% to 80%) on the test set.<br /><br />The researchers concluded that CT texture biomarkers can non-invasively differentiate patients with high and low CD8 lymphocyte infiltration levels. These biomarkers have the potential to serve as surrogate predictors for patient responses to immunotherapy and help in making clinical decisions regarding treatment. The study highlights the need for non-invasive biomarkers to guide clinical decision-making and predict treatment responses in patients, especially in cases where invasive biopsies are not feasible or desirable.

Keywords

CT imaging biomarkers

CD8 lymphocyte infiltration levels

non-small cell lung cancer

NSCLC

immunotherapy

invasive biopsies

RNA-Seq data

radiomics features

logistic regression model

Elastic Net regularization