Multi-parametric MRI-based radiomics for the diagnosis of malignant soft-tissue tumor

Zhibin Yue; Xiaoyu Wang; Tao Yu; Shengjie Shang; Guanyu Liu; Wenwen Jing; Huazhe Yang; Yahong Luo; Xiran Jiang

doi:10.1016/j.mri.2022.05.003

Multi-parametric MRI-based radiomics for the diagnosis of malignant soft-tissue tumor

Magn Reson Imaging. 2022 Sep:91:91-99. doi: 10.1016/j.mri.2022.05.003. Epub 2022 May 4.

Authors

Zhibin Yue¹, Xiaoyu Wang², Tao Yu², Shengjie Shang¹, Guanyu Liu², Wenwen Jing³, Huazhe Yang⁴, Yahong Luo⁵, Xiran Jiang⁶

Affiliations

¹ Department of Biomedical Engineering, School of Intelligent Medicine, China Medical University, Shenyang 110122, PR China.
² Department of Radiology, Cancer Hospital of China Medical University, Liaoning Cancer Hospital and Institute, Shenyang 110042, PR China.
³ Department of Medical Microbiology and Parasitology, Shanghai Medical College of Fudan University, Shanghai 200032, PR China.
⁴ Department of Biophysics, School of Fundamental Sciences, China Medical University, Shenyang 110122, PR China.
⁵ Department of Radiology, Cancer Hospital of China Medical University, Liaoning Cancer Hospital and Institute, Shenyang 110042, PR China. Electronic address: luoyahong8888@hotmail.com.
⁶ Department of Biomedical Engineering, School of Intelligent Medicine, China Medical University, Shenyang 110122, PR China. Electronic address: xrjiang@cmu.edu.cn.

PMID: 35525523
DOI: 10.1016/j.mri.2022.05.003

Abstract

Purpose: To develop and validate a multiparametric magnetic resonance imaging-based radiomics nomogram for differentiating malignant and benign soft-tissue tumors.

Methods: A total of 91 patients with pathologically confirmed soft-tissue tumors were enrolled between January 2017 and October 2020. Forty-eight patients were consecutively enrolled between November 2020 and March 2022, as a time-independent cohort. All patients underwent contrast-enhanced T1-weighted and T2-weighted fat-suppression magnetic resonance scans at 3.0 T. Radiomics features were extracted and selected from the two modalities to develop the radiomics signature. Significant clinical/morphological characteristics were identified using a multivariate logistic regression analysis. The least absolute shrinkage and selection operator regression were applied to identify discriminative features. A clinical-radiomics nomogram was constructed based on clinical/morphological characteristics and radiomics features. Finally, the performance of the nomogram was validated using the receiver operating characteristic and decision curve analysis (DCA).

Results: Six features were selected to establish the combined RS. Size, margin, and peritumoral edema were identified as the most important clinical and morphological factors, respectively. The radiomics signature outperformed the clinical model in terms of AUC and sensitivity. The nomogram integrating the combined RS, size, margin, and peritumoral edema achieved favorable predictive efficacy, generating AUCs of 0.954 (95% confidence interval [CI]: 0.907-1.000, Sen = 0.861, Spe = 0.917), 0.962 (95% CI: 0.901-1.000, Sen = 0.944, Spe = 0.923), and 0.935 (95% CI: 0.871-0.998, Sen = 0.815, Spe = 0.952) in the training (n = 60), validation (n = 31) and time-independent (n = 48) cohorts, respectively. The DCA curve indicated good clinical usefulness of the nomogram.

Conclusions: Our study demonstrated the clinical potential of multiparametric MRI-based radiomics in distinguishing malignant from benign soft-tissue tumors, which can be considered as a noninvasive tool for individual treatment management.

Keywords: MRI; Nomogram; Radiomics; Soft-tissue tumor.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Humans
Magnetic Resonance Imaging
Multiparametric Magnetic Resonance Imaging*
Nomograms
Retrospective Studies
Soft Tissue Neoplasms* / diagnostic imaging