AUTHOR=Jiang Yi , Sun Jingjing , Xia Yuwei , Cheng Yan , Xie Linjun , Guo Xia , Guo Yingkun 

TITLE=Preoperative Assessment for Event-Free Survival With Hepatoblastoma in Pediatric Patients by Developing a CT-Based Radiomics Model

JOURNAL=Frontiers in Oncology

VOLUME=Volume 11 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2021.644994

DOI=10.3389/fonc.2021.644994

ISSN=2234-943X

ABSTRACT=Objective: To develop a CT-based radiomics signature to predict EFS (event-free survival) in hepatoblastoma patients and assess its incremental value to updating staging system and clinicopathological risk factors for individual EFS prediction performance.
Patients and Methods: Data from 88 patients (mean age: 2.28 ± 2.72 years) identified with histologically confirmed hepatoblastoma between 2002 and 2019 from two institutions were enrolled in this retrospective study, and all the patients were followed up at least within 10 months. Radiomics features were manually extracted from hepatoblastoma lesions delineated on computed tomography images on portal venous-phase before treatment. The samples were divided into training cohort and test cohort on different center data. On the training cohort, least absolute shrinkage and selection operator (LASSO) Cox regression model was used for feature selection, radiomics signature building, and a radiomics score (Rad-score) was calculated. Association between the radiomics signature and EFS was explored and radiomics nomogram model was generated with these significant factors in the multiple-variate analysis were constructed in the training cohort, and its prognostic accuracy was evaluated in external validation cohort. The performance of the nomogram was assessed with respect to its discrimination and clinical usefulness. For individualized EFS estimation incremental value of the radiomics signature beyond clinical-pathologic risk factors was assessed using concordance index (C-index). 
Results: The radiomics signatures were independently significantly associated with the EFS (hazard ratio [HR], 19.234; p < 0.05). Incorporating the radiomics signature into the radiomics-based nomogram resulted in good performance (p <0 .001) for the estimation of EFS (C-index: 0.88; 95% confidence interval [CI]: 0.625, 0.783), and the radiomics signature showed performance comparable to clinicopathological factors in estimation of EFS (C-index, 0.88 vs 0.81). Radiomics signature to the nomogram failed to show incremental prognostic value to clinicopathological factors. The combined model with radiomics signature and clinicopathological parameters showed significant improvement in the accuracy of prognosis (C-index, 0.976; 95%CI: 0.652-0.764). 
Conclusion: The radiomics signature is an independent prognostic indicator for EFS in hepatoblastoma patients, providing additional prognostic information beyond clinical-pathologic risk factors. Combination of the radiomics signature and clinicopathological risk factors performed better for individualized EFS prediction with hepatoblastoma patients.