Preoperative computed tomography‐guided disease‐free survival prediction in gastric cancer: a multicenter radiomics study

Abstract

Preoperative and noninvasive prognosis evaluation remains challenging for gastric cancer. Novel preoperative prognostic biomarkers should be investigated. This study aimed to develop multidetector-row computed tomography (MDCT)-guided prognostic models to direct follow-up strategy and improve prognosis. Methods: A retrospective dataset of 353 gastric cancer patients were enrolled from two centers and allocated to three cohorts: training cohort (n = 166), internal validation cohort (n = 83), and external validation cohort (n = 104). Quantitative radiomic features were extracted from MDCT images. The least absolute shrinkage and selection operator penalized Cox regression was adopted to construct a radiomic signature. A radiomic nomogram was established by integrating the radiomic signature and significant clinical risk factors. We also built a preoperative tumor-node-metastasis staging model for comparison. All models were evaluated considering the abilities of risk stratification, discrimination, calibration, and clinical use. Results: In the two validation cohorts, the established four-feature radiomic signature showed robust risk stratification power (P = 0.0260 and 0.0003, log-rank test). The radiomic nomogram incorporated radiomic signature, extramural vessel invasion, clinical T stage, and clinical N stage, outperforming all the other models (concordance index = 0.720 and 0.727) with good calibration and decision benefits. Also, the 2-yr disease-free survival (DFS) prediction was most effective (time-dependent area under curve = 0.771 and 0.765). Moreover, subgroup analysis indicated that the radiomic signature was more sensitive in risk stratifying patients with advanced clinical T/N stage. Conclusions: The proposed MDCT-guided radiomic signature was verified as a prognostic factor for gastric cancer. The radiomic nomogram was a noninvasive auxiliary model for preoperative individualized DFS prediction, holding potential in promoting treatment strategy and clinical prognosis.

Intratumoral and peritumoral radiomics analysis for preoperative Lauren classification in gastric cancer

“…Radiomics utilizes automated quantitative characterization algorithms to transform a large number of excavatable spatial ROI-based image data into representative and effective radiomic features [ 6 ]. Recent advancements in radiomics have provided new ideas for individualized management of GC, including lymphatic metastasis prediction [ 7 , 8 ], distant metastasis prediction [ 9 ], therapeutic response evaluation [ 10 ], and prognostic evaluation [ 11 , 12 ]. These studies highlighted the value of radiomics, suggesting that radiomics could be a potential tool for the Lauren classification in GC.…”

Endoscopic Ultrasound vs. Computed Tomography for Gastric Cancer Staging: A Network Meta-Analysis

“…For example, MDCTbased texture features such as the apparent diffusion coefficient seem to be promising biomarkers for the evaluation of the aggressiveness (T and N stage), treatment response and prognosis of gastric cancer [30]. Furthermore, MDCT imaging biomarkers hold promise as prognostic factors, with potential for guiding treatment and follow-up strategy [31,32].…”

Artificial Intelligence in the Imaging of Gastric Cancer: Current Applications and Future Direction

“…Prognosis is the research highlight of AI-based studies and numerous researchers have explored the potential of hand-crafted radiomics and DL features for prognosis prediction ( 10 , 18 , 20 , 23 , 25 , 30 , 33 , 34 , 43 – 46 , 53 , 60 ). Nine studies directly correlated hand-crafted radiomics and DL features with prognosis ( 10 , 18 , 20 , 23 , 30 , 44 , 46 , 53 , 60 ), and five constructed AI-based models to predict certain clinicopathological features, which were shown to be related to prognosis ( 25 , 33 , 34 , 43 , 45 ). Only one study extracted hand-crafted radiomics features from PET images ( 23 ), whereas all others used CT images ( 10 , 13 , 18 – 22 , 25 , 30 , 33 , 34 , 43 – 46 , 53 , 60 ).…”

“…Nine studies directly correlated hand-crafted radiomics and DL features with prognosis ( 10 , 18 , 20 , 23 , 30 , 44 , 46 , 53 , 60 ), and five constructed AI-based models to predict certain clinicopathological features, which were shown to be related to prognosis ( 25 , 33 , 34 , 43 , 45 ). Only one study extracted hand-crafted radiomics features from PET images ( 23 ), whereas all others used CT images ( 10 , 13 , 18 – 22 , 25 , 30 , 33 , 34 , 43 – 46 , 53 , 60 ). Earlier studies reported OS-related hand-crafted radiomics features ( 10 , 18 ), with later studies building hand-crafted radiomics and DL models that integrated hand-crafted radiomics features with and without clinicopathological features; these achieved good performance in OS, disease-free survival (DFS), and early recurrence prediction ( 20 , 23 , 30 , 33 , 44 ).…”