# CT-based radiomics for prediction of response to neoadjuvant immunochemotherapy in patients with esophageal carcinoma

## TL;DR

This study uses CT-based radiomics to predict how well esophageal cancer patients will respond to a specific type of pre-surgery treatment.

## Contribution

A novel radiomic model combining intra-tumoral and peri-tumoral features for predicting response to neoadjuvant immunochemotherapy in ESCC patients.

## Key findings

- The Rad-Score model achieved AUC values of 0.838 in training and 0.769 in external validation for predicting treatment response.
- The nomogram model showed the best performance with AUC values of 0.867 in training and 0.818 in external validation.
- Radiomic features from both tumor and surrounding regions improved predictive accuracy for treatment response.

## Abstract

In order to investigate the value of radiomic features derived from enhanced computed tomography (CT) for assessment of therapeutic efficacy in patients with Esophageal squamous cell carcinoma (ESCC) underwent neoadjuvant immunochemotherapy (NICT).

The primary cohort of this study included 120 ESCC patients who received NICT from April 2020 to August 2023, comprising 52 patients with good responders (GR) and 68 patients with non-good responders (non-GR) after NICT, the external validation cohort included 30 patients from another hospital, comprising 14 patients with GR and 16 patients with non-GR after NICT. Features were derived from both the intra-tumoral and peri-tumoral regions of the tumor in the enhanced CT image, and the least absolute shrinkage and selection operator (LASSO) regression was used to establish predictive radiomic models (Rad-Scores) and T-stage model for predicting therapeutic response to NICT.

The Rad-Score for predicting response to NICT generated the area under the curve (AUC) values of 0.838, 0.831, and 0.769 in the training, internal validation, and external validation cohorts, respectively. For T-stage, corresponding AUC values were 0.809, 0.800, and 0.716 in the same cohorts. Additionally, the nomogram model produced AUC values of 0.867, 0.871, and 0.818 in the training, internal validation, and external validation cohorts, respectively.

The established models demonstrate promising predictive potential for assessing the efficacy of NICT in ESCC patients, which may assist clinicians in formulating appropriate treatment strategies.