Paper Proffered Program Diagnostic and Interventional Radiology Physics

Multi-Sequence MRI-Derived Radiomics and Machine Learning for the Stratification of Suspected Recurrent Brain Metastases

Abstract
Purpose

To investigate the diagnostic performance of perfusion and diffusion MRI in differentiating progressing tumours from radiation necrosis in brain metastases (BM) post-radiosurgery, using a radiomic-based machine learning classification approach.

Methods

Imaging data were collected from patients with confirmed TP/RN diagnoses enrolled in an ongoing clinical trial. The resulting dataset (n = 55) comprised diffusion, perfusion, and permeability MRI sequences (DWI-ADC, DSC-MRI, and DCE-MRI) in addition to standard anatomical MRI acquisitions (pre- and post-contrast T1-weighted MRI and post-contrast T2-weighted MRI), including BM across a broad range of lesion volumes (20,000 mm³). Images were acquired across two different sites, using three 1.5T-3T SIEMENS scanners (Biograph mMR, Magnetom Vida, Magnetom Sola). BM contours were automatically interpolated to subdivide the clinical gross tumour volume into sub-compartments corresponding to the lesion core, edge, and periphery. Radiomic features capturing intensity-based and textural characteristics were extracted for each compartment and sequence. Feature selection was performed to remove highly inter-correlated redundant features; the resulting radiomic datasets were used to train sequence- and compartment-specific Random Forest classifiers. Model performance was assessed using leave-one-out cross-validation.

Results

The best-performing classifier incorporated DSC-MRI cerebral blood flow-derived features extracted within the lesion edge (ROC-AUC: 0.829; 95% CI: 0.719–0.921; sensitivity: 0.760; specificity: 0.750), surpassing the corresponding negative radiomic control using solely shape-based features (ROC-AUC: 0.604; 95% CI: 0.470–0.733; sensitivity: 0.733; specificity: 0.630) as well as the T1-weighted post-contrast clinical reference classifier within the same region (ROC-AUC: 0.596, 95% CI: 0.437–0.700; sensitivity: 0.700; specificity: 0.481).

Conclusion

Our study offers a unique comparison of multiple co-acquired MRI modalities for the classification of suspected BM recurrence. Preliminary findings indicate that perfusion MRI–derived imaging biomarkers can support progression/necrosis case stratification preferentially over conventional anatomical MRI, consistent with evolving clinical practice within which perfusion MRI is increasingly adopted as part of standard-of-care for post-treatment BM management.

People

Related

Similar sessions

Poster Poster Program
Jul 19 · 07:00
B-Trac – Breast Tissue Rotation and Compression Apparatus for Calibration

Mammography (compressed 2D) and MRI (uncompressed 3D) capture breast tissue under different conditions, complicating tumor localization across modalities. To bridge this gap, we developed a customizable physical platform to simul...

Dayadna Hernandez Perez
Diagnostic and Interventional Radiology Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
Comprehensive Medical Physics Assessment of Digital Mammography Equipment: A Three-Year Multi-Site Evaluation of Technical Performance and Radiation Safety at 24 Saudi Arabian Healthcare Institutions (2022–2024)

To conduct a comprehensive multi-center audit evaluating the technical performance, image quality, and radiation safety of digital mammography systems across 24 unique healthcare facilities in Saudi Arabia. This study aims to est...

Sami Alshaikh, PhD
Diagnostic and Interventional Radiology Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
Starting Small: Implementing a CT Protocol Optimization Program

This talk describes our organization’s CT optimization program, and how we implemented it to make efficient use of limited physicist time.

Robert J. Cropp, PhD
Diagnostic and Interventional Radiology Physics 0 people interested