TY - JOUR
T1 - Prediction of neoadjuvant chemotherapy response in high-grade osteosarcoma
T2 - Added value of non-tumorous bone radiomics using CT images
AU - Xu, Lei
AU - Yang, Pengfei
AU - Hu, Kun
AU - Wu, Yan
AU - Xu-Welliver, Meng
AU - Wan, Yidong
AU - Luo, Chen
AU - Wang, Jing
AU - Wang, Jinhua
AU - Qin, Jiale
AU - Rong, Yi
AU - Niu, Tianye
N1 - Funding Information:
Funding: This work was supported by National Key R&D Program of China (2018YFE0114800), Natural Science Foundation of China (NSFC Grant No. 81871351, 81950410632), Zhejiang Postdoctoral Scientific Research Preferred Project (zj2019004).
Publisher Copyright:
© Quantitative Imaging in Medicine and Surgery. All rights reserved.
PY - 2021/4
Y1 - 2021/4
N2 - Background: This study aimed to determine the impact of including radiomics analysis of non-tumorous bone region of interest in improving the performance of pathological response prediction to chemotherapy in high-grade osteosarcomas (HOS), compared to radiomics analysis of tumor region alone. Methods: This retrospective study included 157 patients diagnosed with HOS between November 2013 and November 2017 (age range, 5–44 years; mean age, 16.99 ±7.42 years), in which 69 and 88 patients were diagnosed as pathological good response (pGR) and non-pGR, respectively. Radiomics features were extracted from tumor and non-tumorous bone regions based on diagnostic CT images. Pathological response classifiers were developed and validated via leave-one-out cross validation (LOOCV) and independent validation methods by using the area under the receiver operating characteristic curve (AUC) value as the figure of merit. Results: Using the LOOCV, the classifiers combining features from tumor and non-tumorous regions showed better prediction performance than those from tumor region alone (AUC, 0.8207±0.0043 vs. 0.7799±0.0044). The combined classifier also showed better performance than the tumor feature-based classifier in both training and validation datasets [training dataset: 0.791, 95% confidence interval (CI), 0.706–0.860 vs. 0.766, 95% CI, 0.679–0.840; validation dataset: 0.816, 95% CI, 0.662–0.920 vs. 0.766, 95% CI, 0.606–0.885]. Conclusions: Radiomics analysis of combined tumor and non-tumorous bone features showed improved performance of pathological response prediction to chemotherapy in HOS compared to that of tumor features alone. Moreover, the proposed classifier had the potential to predict pathological response to chemotherapy for HOS patients.
AB - Background: This study aimed to determine the impact of including radiomics analysis of non-tumorous bone region of interest in improving the performance of pathological response prediction to chemotherapy in high-grade osteosarcomas (HOS), compared to radiomics analysis of tumor region alone. Methods: This retrospective study included 157 patients diagnosed with HOS between November 2013 and November 2017 (age range, 5–44 years; mean age, 16.99 ±7.42 years), in which 69 and 88 patients were diagnosed as pathological good response (pGR) and non-pGR, respectively. Radiomics features were extracted from tumor and non-tumorous bone regions based on diagnostic CT images. Pathological response classifiers were developed and validated via leave-one-out cross validation (LOOCV) and independent validation methods by using the area under the receiver operating characteristic curve (AUC) value as the figure of merit. Results: Using the LOOCV, the classifiers combining features from tumor and non-tumorous regions showed better prediction performance than those from tumor region alone (AUC, 0.8207±0.0043 vs. 0.7799±0.0044). The combined classifier also showed better performance than the tumor feature-based classifier in both training and validation datasets [training dataset: 0.791, 95% confidence interval (CI), 0.706–0.860 vs. 0.766, 95% CI, 0.679–0.840; validation dataset: 0.816, 95% CI, 0.662–0.920 vs. 0.766, 95% CI, 0.606–0.885]. Conclusions: Radiomics analysis of combined tumor and non-tumorous bone features showed improved performance of pathological response prediction to chemotherapy in HOS compared to that of tumor features alone. Moreover, the proposed classifier had the potential to predict pathological response to chemotherapy for HOS patients.
KW - Added value
KW - High-grade osteosarcoma (HOS)
KW - Neoadjuvant chemotherapy response
KW - Non-tumorous bone features
UR - http://www.scopus.com/inward/record.url?scp=85101300851&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85101300851&partnerID=8YFLogxK
U2 - 10.21037/qims-20-681
DO - 10.21037/qims-20-681
M3 - Article
AN - SCOPUS:85101300851
VL - 11
SP - 1184
EP - 1195
JO - Quantitative Imaging in Medicine and Surgery
JF - Quantitative Imaging in Medicine and Surgery
SN - 2223-4292
IS - 4
ER -