# Radiomic features and tumor immune microenvironment associated with anaplastic lymphoma kinase-rearranged lung adenocarcinoma and their prognostic value

## TL;DR

This study shows how CT image features and immune markers can help identify a specific lung cancer type and predict patient outcomes.

## Contribution

The study introduces a combined radiomic and immune-based model for identifying ALK-rearranged lung adenocarcinoma and predicting prognosis.

## Key findings

- HLA-I and PD-L1 expression were significantly lower in ALK-rearranged tumors.
- The combined radiomic and clinicopathological model outperformed individual models in predicting ALK rearrangements.
- The combined model also improved DFS prediction for ALK-rearranged patients compared to clinicopathological models alone.

## Abstract

To identify radiomic features from preoperative computed tomography (CT) images and characteristics of the tumor immune microenvironment (TIME) associated with anaplastic lymphoma kinase (ALK) rearrangement in patients with lung adenocarcinomas and their prognostic value in predicting recurrence or metastases after surgery.

This retrospective study included 66 ALK-positive and 66 ALK-negative patients who underwent surgical resected lung adenocarcinoma. The number of CD8+ T cells and Human leukocyte antigen class I (HLA-I)/programmed death ligand 1 (PD-L1) expression were determined using immunohistochemistry. Radiomic features were extracted from the preoperative CT images. Combined radiomic, clinicopathological, and clinicopathological-radiomic models were built to predict ALK rearrangements. The models’ prediction performance was analyzed using receiver operating characteristic (ROC) curves with five-fold cross-validation. Prediction models for determining disease-free survival (DFS) of ALK-rearranged patients were developed, and the C-index after internal cross-validation was calculated to evaluate the performance of the models.

HLA-I and PD-L1 expression were negatively associated with ALK rearrangement (both P < 0.001). The ROC curve indicated areas under the curve of 0.763, 0.817, and 0.878 for the radiomics, clinicopathology, and combined models in predicting ALK rearrangement, respectively. The combined model showed significant improvement compared to the clinicopathological (P = 0.02) and radiomics (P < 0.001) models alone. The validation C-indices were 0.752, 0.712, and 0.808 for the radiomic, clinicopathological, and combined models in predicting the DFS of ALK-rearranged patients, respectively. The combined model showed a significant improvement (P < 0.001) compared to the clinicopathological model alone.

This study demonstrated the potential role of radiomics and TIME characteristics in identifying ALK rearrangements in lung adenocarcinomas and the prognostic value of radiomics in predicting DFS in patients with ALK rearrangements.