Page 35 - ARNM-3-2

Advances in Radiotherapy
& Nuclear Medicine Radiomics for gastric cancer

Borrmann type IV GC (40 cases) from primary gastric aiding treatment decision-making, providing an indication
lymphoma (30 cases) with 485 radiomic features, achieving of prognosis, and assisting the evaluation of treatment
a diagnosis accuracy of 81.43%, 84.29%, and 87.14% outcomes. Clinically, accurate TNM staging before
33
for the subjective findings model, radiomic signature, surgery is difficult and usually performed using invasive
and combined model (radiomic signature, subjective methods. The emergence of radiomics provided a non-
34
CT findings, age, and gender), respectively. Sun et al. invasive way for TNM classification preoperatively for
21
assessed the texture analysis of PET in the differentiation patients with GC, as shown in Table 2.
of GC (45 cases) and gastric lymphoma (34 cases). The Chen et al. developed a diffusion-weighted imaging-
35
result demonstrated that inertia GLCM and entropy
GLCM were the most discriminating features with an area based radiomic nomogram for LNM prediction in 146
advanced GC patients and achieved an AUC of 0.850,
under the curve (AUC) of 0.714 and 0.770, respectively, 0.857, and 0.878 in the training, internal validation, and
for differentiating gastric lymphoma from GC and low- external validation, respectively. CT-based radiomic studies
grade gastric lymphoma from GC. Sun et al. investigated
22
the contrast-enhanced CT (CECT)-based radiomics were mainly focused on the pre-operative prediction of
38
36-46
in differentiating GC (60 cases) from gastric stromal LNM. Wang et al. investigated the value of CT-based
tumors (40 cases), and the result indicated that the model radiomics in the differentiation between T2 and T3/4 stage
integrating subjective CT signs and radiomic signature lesions of 244 patients with GC and achieved an AUC of
achieved the highest AUC (0.903), specificity (93.33%), 0.843 (95% CI = 0.746 – 0.914) and 0.818 (95% CI = 0.711
and accuracy (86.00%). In a study conducted by Wang – 0.899) in the training and test cohort, respectively.
42
et al., a CT-based radiomic nomogram was developed to Yardımcı et al. evaluated the feasibility and accuracy of
23
differentiate gastric neuroendocrine carcinomas (63 cases) CT texture analysis in the differentiation between T stages
from gastric adenocarcinomas (ADCs) (63 cases). The (T1 and T2 vs. T3 and T4), N stages (N+ vs. N–), and grades
nomogram achieved an AUC of 0.821 (95% confidence (low-intermediate vs. high) in 114 GC patients, achieving
interval [CI] = 0.725 – 0.895) in the primary cohort and discriminatory capacities of 90.4%, 81.6% and 64.5% for
0.809 (95% CI = 0.649 – 0.918) in the validation cohort T stage, N stage, and tumor grade, respectively. PET-CT-
by integrating the radiomic signature, tumor margin, and based radiomics was also reported for the prediction of
lymph node metastasis (LNM) into the model. N2-3b metastasis, LNM, and N stage with promising
accuracy. 47,48
The histological grade of GCs plays an important
role in selecting the treatment methods and predicting Lymphatic vascular invasion (LVI) and perineural
the prognosis. 24,25 Zhang et al. and Li et al. 26,27 applied invasion (PNI) are two important prognostic factors
49
apparent diffusion coefficient maps of MRI and CECT for the treatment outcome of patients with GC. Chen
50
images to predict the histopathological grades of et al. applied CECT-based radiomics to predict LVI and
87 patients and 554 patients with GCs, respectively. Both clinical outcome preoperatively in 160 GC patients and
studies demonstrated good performance in successfully demonstrated that radiomic features may serve as potential
differentiating between the histological grades of GC. The markers for the prediction of LVI and progression-free
51
Lauren classification is one of the mainstream histological survival for GC patients. Yardımcı et al. indicated that CT
classification methods for GC, which plays a very important texture analysis has the potential to predict LVI and PNI in
52
role in treatment decision-making and prognosis tubular gastric ADC using machine learning. Zheng et al.
evaluation. 28,29 Wang et al. developed a nomogram based demonstrated that CECT-based radiomics is feasible for
30
on CT radiomics to differentiate the Lauren diffuse type the non-invasive pre-operative prediction of PNI in GC.
from the intestinal type preoperatively in 539 GC patients PET-CT-based radiomics was investigated in two studies
and achieved a validation AUC of 0.758. A study by Wang for LVI prediction, with 101 and 148 enrolled GC cases,
et al. demonstrated that multiphase CT radiomic-based achieving the highest AUC of 0.944 and 0.914, respectively,
31
nomogram was promising for a pre-operative distinction in the validation cohorts for the combined models. 53,54
of intestinal-type gastric ADCs. In a multicenter study
32
with 693 GC patients, Chen et al. discovered that a CT 3.4. Prediction of tumor prognosis and treatment
radiomic-based nomogram was able to distinguish the response
diffuse type and signet ring cell carcinoma preoperatively. Although the TNM staging system is clinically the
main prognostic prediction for patients with GC,
3.3. Prediction of TNM stage it is inadequate for prognosis determination due to
The TNM stage classification is the global standard to the inherent heterogeneity of GC. Biomarkers from
55
classify the stage of solid tumors, which is important in advanced molecular biology are promising in prognostic

Volume 3 Issue 2 (2025) 27 doi: 10.36922/arnm.8350