Improving Diagnostic Robustness of Perfusion MRI in Brain Metastases: A Focus on 3D ROI Techniques and Automatic

Stéphanie Rudzinska-Mistarz1, Brieg Dissaux1,2, Laurie Marchi1

  • 1Radiology Department, University Hospital, 29200 Brest, France.

Cancers
|July 12, 2025
PubMed

Insights

A novel 3D region of interest (ROI) method for perfusion MRI shows improved accuracy in distinguishing brain tumor recurrence from radiation necrosis, outperforming traditional manual ROI techniques.

Area of Science:

  • Neuro-oncology
  • Radiology
  • Medical Imaging

Background:

  • Differentiating tumor recurrence from radiation necrosis post-radiotherapy for brain metastases is diagnostically challenging.
  • Perfusion MRI, specifically relative cerebral blood volume (rCBV) measurement, is utilized but sensitive to region of interest (ROI) placement variability.
  • Accurate differentiation is crucial for appropriate patient management and treatment planning.

Purpose of the Study:

  • To compare the diagnostic performance of different cerebral perfusion MRI methods, including a novel volumetric 3D ROI approach and automatic thresholding.
  • To evaluate the effectiveness of these methods in distinguishing between tumor recurrence and radiation necrosis.
  • To identify an optimal rCBV threshold for differentiating relapse from necrosis.

Main Methods:

  • Retrospective analysis of 23 patients with 25 brain metastases treated with stereotactic radiotherapy.
  • Perfusion MRI was performed before biopsy or surgical resection for histological confirmation.
  • Diagnostic performance was assessed using area under the ROC curve (AUC), sensitivity, and specificity for manual and 3D ROI methods, alongside an automatic thresholding technique.

Main Results:

  • The 3D ROI method, incorporating the entire lesion and a healthy caudate nucleus ROI, achieved superior diagnostic performance with an AUC of 0.65 compared to manual methods (AUC = 0.53).
  • Moderate robustness was observed, with an intraclass correlation coefficient of 0.60 between different software platforms.
  • Automatic thresholding generated tumor sub-volumes to identify specific rCBV cut-off values.

Conclusions:

  • The volumetric 3D ROI method demonstrates potential for enhancing diagnostic accuracy in differentiating tumor recurrence from radiation necrosis.
  • Further validation with standardized protocols and larger patient cohorts is recommended.
  • This technique may offer a more reliable approach to perfusion MRI analysis in neuro-oncology.