Skip to main navigation Skip to search Skip to main content

Radiomics-based Differentiation of Recurrent Brain Metastases from Treatment Effects: A Multi-Institutional Comparative Study with Advanced Imaging

  • Hyemin Um
  • , Marwa Ismail
  • , Virginia B. Hill
  • , Sushant Puri
  • , Jennifer S. Yu
  • , Lan Lu
  • , Ameya P. Nayate
  • , Anthony Higinbotham
  • , Lisa R. Rogers
  • , Prateek Prasanna
  • , Mainak Bardhan
  • , Chengnan Li
  • , Mustafa M. Basree
  • , Andrew M. Baschnagel
  • , Alan B. McMillan
  • , Ankush Bhatia
  • , Manmeet S. Ahluwalia
  • , Michael C. Veronesi
  • , Pallavi Tiwari
  • University of Wisconsin-Madison
  • Northwestern University
  • Oregon Health and Science University
  • Johns Hopkins University
  • Cleveland Clinic Foundation
  • Case Western Reserve University
  • University of Virginia
  • Mich
  • Baptist Hospital Miami
  • Department of Veterans Affairs

Research output: Contribution to journalArticlepeer-review

Abstract

Purpose To compare the ability of radiomics, advanced imaging modalities, and Response Assessment in Neuro-Oncology Brain Metastases (RANO-BM) criteria to distinguish tumor recurrence from treatment effects in patients with brain metastasis. Materials and Methods Posttreatment MRI (gadolinium-enhanced T1-weighted, T2-weighted, fluid-attenuated inversion recovery [FLAIR]) data of patients from three institutions, obtained between January 2001 and March 2023, were retrospectively analyzed. Reference standards were established by histopathology or according to clinical and imaging courses. Following preprocessing and segmenting tumors into subcompartments of enhancing lesion, edema, or necrotic core, 1104 radiomic features were extracted, and classification was conducted using random forest models. A comparative assessment with advanced imaging and RANO-BM was performed using a subset of studies with available modalities (dynamic susceptibility contrast MRI, fluorine 18 [18F] fluorodeoxyglucose [FDG] PET/MRI, 18F-FDG PET/CT, and longitudinal gadolinium-enhanced T1-weighted MRI). Statistical analyses were performed using the McNemar test. Results The study included data from 242 patients (mean age, 59 years ± 11.5; 114 female patients). Models including T2-weighted MRI features from the enhancing lesion, FLAIR features from edema, and FLAIR features from necrotic core subcompartments yielded accuracies of 70.6%, 71.4%, and 72.0%, respectively. The average accuracies for the enhancing lesion, edema, and necrotic subcompartments across different study sites were 68.9%, 67.9%, and 70.7%, respectively. In comparative subset analyses, radiomics outperformed other techniques in distinguishing tumor recurrence from treatment effects (eg, 76.5% vs 39.2% accuracy using RANO-BM, P < .001). Conclusion Radiomics outperformed RANO-BM criteria and advanced imaging modalities in differentiating tumor recurrence from treatment effects, with consistently higher accuracies across lesion subcompartments. Keywords: Brain Metastases, Treatment Effects, Radiomics, RANO-BM, Advanced Imaging Supplemental material is available for this article.

Original languageEnglish
Pages (from-to)e250435
JournalRadiology: Imaging Cancer
Volume8
Issue number3
DOIs
StatePublished - May 1 2026

Keywords

  • Advanced Imaging
  • Brain Metastases
  • Radiomics
  • RANO-BM
  • Treatment Effects

Fingerprint

Dive into the research topics of 'Radiomics-based Differentiation of Recurrent Brain Metastases from Treatment Effects: A Multi-Institutional Comparative Study with Advanced Imaging'. Together they form a unique fingerprint.

Cite this