Applicability of Radiomics for Differentiation of Pancreatic Adenocarcinoma from Healthy Tissue of Pancreas by Using Magnetic Resonance Imaging and Machine Learning DOI Open Access
Dimitrije Šarac, Milica Badža Atanasijević, Milica Mitrović-Jovanović

et al.

Cancers, Journal Year: 2025, Volume and Issue: 17(7), P. 1119 - 1119

Published: March 27, 2025

This study analyzed different classifier models for differentiating pancreatic adenocarcinoma from surrounding healthy tissue based on radiomic analysis of magnetic resonance (MR) images. We observed T2W-FS and ADC images obtained by 1.5T-MR 87 patients with histologically proven training validation purposes then tested the most accurate predictive that were another group 58 patients. The tumor segmented three consecutive slices, largest area interest (ROI) marked using MaZda v4.6 software. resulted in a total 261 ROIs each classes training-validation 174 testing group. software extracted 304 features ROI, divided into six categories. was conducted through feature reduction methods five-fold subject-wise cross-validation. In-depth shows best results Random Forest (RF) Mutual Information score (all nine are co-occurrence matrix): an accuracy 0.94/0.98, sensitivity specificity F1-score 0.94/0.98 achieved T2W-FS/ADC group, retrospectively. In 0.69/0.81, 0.86/0.82, 0.52/0.70, 0.74/0.83 images, machine learning approach radiomics relatively high differentiation tissue, which could be especially applicable screening purposes.

Language: Английский

Applicability of Radiomics for Differentiation of Pancreatic Adenocarcinoma from Healthy Tissue of Pancreas by Using Magnetic Resonance Imaging and Machine Learning DOI Open Access
Dimitrije Šarac, Milica Badža Atanasijević, Milica Mitrović-Jovanović

et al.

Cancers, Journal Year: 2025, Volume and Issue: 17(7), P. 1119 - 1119

Published: March 27, 2025

This study analyzed different classifier models for differentiating pancreatic adenocarcinoma from surrounding healthy tissue based on radiomic analysis of magnetic resonance (MR) images. We observed T2W-FS and ADC images obtained by 1.5T-MR 87 patients with histologically proven training validation purposes then tested the most accurate predictive that were another group 58 patients. The tumor segmented three consecutive slices, largest area interest (ROI) marked using MaZda v4.6 software. resulted in a total 261 ROIs each classes training-validation 174 testing group. software extracted 304 features ROI, divided into six categories. was conducted through feature reduction methods five-fold subject-wise cross-validation. In-depth shows best results Random Forest (RF) Mutual Information score (all nine are co-occurrence matrix): an accuracy 0.94/0.98, sensitivity specificity F1-score 0.94/0.98 achieved T2W-FS/ADC group, retrospectively. In 0.69/0.81, 0.86/0.82, 0.52/0.70, 0.74/0.83 images, machine learning approach radiomics relatively high differentiation tissue, which could be especially applicable screening purposes.

Language: Английский

Citations

0