Page 33 - ARNM-3-2

Advances in Radiotherapy
& Nuclear Medicine Radiomics for gastric cancer

each year. Gastrectomy with adequate lymphadenectomy various cancers. By extracting a great number of two-
1,2
12
is still the main treatment for GC. However, even with dimensional and high-dimensional features from images
3
radical resection, the overall survival (OS) and disease- of ultrasound, CT, and MRI using data-characterization
free survival (DFS) remain unsatisfactory, with half of algorithms, radiomics is able to analyze tumor heterogeneity
GC patients relapsing after surgery. In locally advanced and other characteristics non-invasively. 13,14 Radiomic
4,5
cases, the 5-year survival rate is <30%. Therefore, early features have been investigated alone or in combination
6
detection using biomarkers plays a key role in the optimal with other parameters such as pathological, physiological,
management of GC. and genomic data for the classification, diagnosis, clinical
staging, pathological evaluation, and therapeutic response
As early as 1994, the World Health Organization of patients with GC. 15-17 The purpose of this study is to
classified Helicobacter pylori as a Class I carcinogen, systematically review the current evidence, clinical value,
which is closely related to the dysplasia of cancer. In and future potential of radiomics in the diagnosis, clinical
7
addition, biomarkers such as a carcinoembryonic antigen, staging, and prognostic prediction for patients with GC.
carbohydrate antigen (CA) 19-9, CA72-4, CA12-5, and
alpha-fetoprotein were investigated intensively in the 2. Methodology
early diagnosis and staging of gastric tumors in the clinic.
However, most of the identified biomarkers failed in the 2.1. Data sources
validation studies, and the accuracy of a single biomarker PubMed was used to conduct a search for articles related
remains to be verified. The American Joint Committee to radiomics for GC using keywords such as “radiomics,”
8,9
th
on Cancer (8 Edition) suggested that the classification “machine learning,” “gastric cancer,” “gastric carcinoma,”
of the tumor-node-metastasis (TNM) stage for GC is “predict,” and “prognosis.” The articles were selected
based on computed tomography (CT) and endoscopic according to the following recorded information of each
ultrasound, while the evaluation of metastasis is based qualified article: title, publication year, sample size, sample
on the magnetic resonance imaging (MRI) and positron condition, research purpose, imaging modality, research
emission tomography-CT (PET-CT). Despite the results, and more. This selection process was carried out
10
application of various imaging techniques, the diagnostic by two reviewers (ZH and ZQ), and the selected articles
accuracy based on captured image features and visible underwent a second screening by a third reviewer (JX).
abnormalities identified through manual interpretation A total of 52 papers were selected from PubMed (until
remains limited. Therefore, more effective early diagnosis October 2022), as shown in Figure 1. The papers were
11
and prognosis evaluation methods are urgently needed. divided into four categories according to the research
In the last decade, along with the development of direction: diagnostic classification of GC (n = 9), prediction
artificial intelligence, radiomics has been investigated of TNM stages (n = 19), prognosis and response prediction
intensively for the diagnosis and prognostic prediction of to treatment (n = 20), and studies with deep learning

Figure 1. Flow chart of data selection
Abbreviation: TNM: Tumor-node-metastasis.

Volume 3 Issue 2 (2025) 25 doi: 10.36922/arnm.8350