Radiomic features of pretreatment MRI could identify T stage in patients with rectal cancer: Preliminary findings

Yiqun Sun; Panpan Hu; Jiazhou Wang; Lijun Shen; Fan Xia; Gan Qing; Weigang Hu; Zhen Zhang; Chao Xin; Weijun Peng; Tong Tong; Yajia Gu

doi:10.1002/jmri.25969

Radiomic features of pretreatment MRI could identify T stage in patients with rectal cancer: Preliminary findings

J Magn Reson Imaging. 2018 Feb 13. doi: 10.1002/jmri.25969. Online ahead of print.

Authors

Yiqun Sun^{1

2}, Panpan Hu³, Jiazhou Wang³, Lijun Shen³, Fan Xia³, Gan Qing³, Weigang Hu³, Zhen Zhang³, Chao Xin¹, Weijun Peng¹, Tong Tong¹, Yajia Gu¹

Affiliations

¹ Department of Radiology, Fudan University Shanghai Cancer Center, Shanghai, China, Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, China.
² Shanghai Institute of Medical Imaging, Fudan University, Shanghai, China.
³ Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, China.

PMID: 29437279
DOI: 10.1002/jmri.25969

Abstract

Background: Recent studies have shown that magnetic resonance (MR) radiomic analysis is feasible and has some value in identifying tumor characteristics, but there are few data regarding the role of MR-based radiomic features in rectal cancer.

Purpose: The aim of this study was to determine whether radiomic features extracted from T₂ -weighted imaging (T₂ WI) can identify pathological features in rectal cancer.

Study type: Retrospective study.

Population/subjects: A cohort comprising 119 rectal cancer patients who underwent surgery between January 2015 and November 2016.

Field strength/sequence: 3.0T, axial high-resolution T₂ -weighted turbo spin echo (TSE) sequence.

Assessment: Patients were classified according to pathological features such as T stage, N stage, perineural invasion, histological grade, lymph-vascular invasion, tumor deposits, and circumferential resection margin (CRM). The whole tumor volume (WTV) was distinguished, and segments were quantified on axial high-resolution T₂ WI by a radiologist. A total of 256 radiomic features were extracted.

Statistical tests: To achieve reliable results, cluster analysis and least absolute shrinkage and selection operator (LASSO) were implemented. In the cluster analysis, the patients were divided into two groups, and chi-square tests were performed to investigate the relationship between the pathological features and the radiomic-based clusters. The area under the curve (AUC) was calculated to evaluate the predictability of the model in the LASSO analysis.

Results: The cluster results revealed that patients could be stratified into two groups, and the chi-square test results indicated that the pT stage was correlated with the radiomic feature cluster results (P = 0.002). The prediction model AUC for the diagnostic T stage was 0.852 (95% confidence interval: 0.677-1; sensitivity: 79.0%, specificity: 82.0%).

Data conclusion: The use of MRI-derived radiomic features to identify the T stage is feasible in rectal cancer.

Level of evidence: 3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018.

Keywords: feasibility; magnetic resonance imaging; radiomics; rectal cancer.