Year: 2025
Type: article
Study design: Observational Study
Abstract: BACKGROUND Radiomics is a quantitative approach to medical imaging that can improve existing clinical data through advanced mathematical analysis. The aim of the study was to develop a machine learning model for identifying tumor invasion zones in pediatric brain MRI images. MATERIAL AND METHODS Retrospective data were analyzed from children (mean age 7 years) with malignant CNS tumors (n=30, 20 glioblastoma, 10 ATRT) and controls (n=30, diffuse astrocytoma, G2). We preprocessed MRI images and extracted radiomic characteristics. A correlation analysis of radiomic parameters with a histological tumor marker (Ki-67) and the intensity of the heat shock protein Hsp70 was performed. RESULTS Classifier models (Support Vector Machine, Logistic Regression, XGBoost) and a neural network were trained to identify areas of tumor cell invasion in the PTZ with accuracy of 0,85, with an area under the curve (AUC) of 0,89. A correlation was found between GLRLM class features and Ki-67 marker (Pearson’s r = 0,94; p <0,05) and Hsp70 intensity levels (r = 0,88; p <0,05). CONCLUSION In this study, we identified highly informative characteristics of the peritumoral zone and, based on them, trained classifier models and a neural network. Regions predicted by the model overlapped with recurrence zones on follow-up images.
Source: Neuro-Oncology
Authors: M Shevtsov, A Kim, N Mikhailova
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4414800174
FWCI:
Cites:
Cited by:
Related to: