Euro Radio: An MR radiomics model for predicting disease-free and overall survival in patients with esophageal cancer

According to statistics, the incidence of esophageal cancer (EC) ranks seventh in the world, and the overall mortality ranks sixth
.
Esophageal cancer is one of the most common malignant tumors in China, and more than 90% of esophageal cancers are squamous cell carcinomas (ESCC)
.
Although it has been reported that the overall 5-year survival rate for EC has risen from less than 5% in the 1960s to about 20% in the last 10 years, the survival rate of patients with advanced disease remains unsatisfactory
.

Depending on the stage of the disease, ESCC treatment includes surgery, neoadjuvant chemotherapy, and chemoradiation
.
However, because the clinical manifestations of early ESCC are not obvious, most patients are diagnosed as advanced and unresectable
.
Early detection of tumors can greatly improve the survival rate of ESCC patients
.
In addition, a considerable proportion of patients will experience relapse after treatment
.
Therefore, identifying predictors associated with recurrence can lead to personalized treatment and follow-up strategies that can prolong survival
.

The application of radiomics provides a new assessment method for predicting the prognosis of cancer patients
.
At present, most of the radiomics studies on ESCC are based on CT, and there are more and more radiomics studies based on PET
.
However, preoperative MRI radiomics signatures would be more valuable with higher tissue resolution, but few studies have addressed them
.

A study published in the European Radiology journal extracted radiomic features from MRI images based on 1 mm anisotropy-3D contrast-enhanced StarVIBE sequences
.
Combined with clinical risk factors, an optimal model for predicting the survival of ESCC patients was constructed
.

This study collected ESCC patients at our institution from 2015 to 2017 and randomly assigned them to training and validation groups in a 7:3 ratio
.
Random survival forest (RSF) and variable hunting were used to screen radiomic features, and LASSO-Cox regression analysis was used to establish three models, including clinical-only model, radiomics-only model, and combined clinical-radiomics model.
Sex index (CI) and calibration curve were evaluated
.
Use nomograms and decision curve analysis (DCA) to display intuitive forecast information
.
 

Seven radiomic features were selected from 434 patients and combined with statistically significant clinical features to construct predictive models for disease-free survival (DFS) and overall survival (OS)
.
The integrated model had the highest performance in predicting DFS ([CI], 0.
714, 0.
729) and OS ([CI], 0.
730, 0.
712) in both training and validation sets
.
DCA showed that the combined and clinical models had significantly higher net benefits than the radiomics model alone at different threshold probabilities
.
 

Figure Calibration curves for nomograms across all cohorts
.
A joint model nomination plot to predict the calibration curve for DFS
.
B Joint Model Nomination Plot for Predicting OS Calibration Curve

This study demonstrates that an MRI-based model combining radiomics and clinical factors can improve survival prediction performance in ESCC patients
.

Original source:

Funing Chu, Yun Liu, Qiuping Liu, et al.
Development and validation of MRI-based radiomics signatures models for prediction of disease-free survival and overall survival in patients with esophageal squamous cell carcinoma.
DOI: 10.
1007/s00330-022-08776- 6