Differentiation between high-grade gliomas and solitary brain metastases: a comparison of five diffusion-weighted MRI models

Jiaji Mao, Weike Zeng, Qinyuan Zhang, Zehong Yang, Xu Yan, Huiting Zhang, Mengzhu Wang, Guang Yang, Min-Xiong Zhou, Jun ShenBMC Medical Imaging 2020 · DOI

Introduction

Distinguishing primary brain tumors, such as high-grade gliomas (HGGs), from solitary brain metastases (SBMs) remains a significant clinical challenge. On standard structural MRI, these lesions often appear identical—typically presenting as single, contrast-enhancing masses. However, their clinical management is vastly different: HGGs generally require immediate surgical resection and aggressive adjuvant therapy, whereas SBMs may be managed with stereotactic radiosurgery or systemic treatments depending on the primary cancer.

Currently, standard MRI fails to differentiate these entities in over 40% of cases, often leading to invasive brain biopsies. This study investigates whether advanced diffusion-weighted imaging (DWI) models can provide a non-invasive "microscopic fingerprint" to accurately distinguish HGGs from SBMs.

Methods and Study Population

The researchers prospectively recruited 41 patients (20 with HGGs and 21 with SBMs) who underwent high-field (3.0T) diffusion MRI. Five mathematical models of water diffusion, ranging from standard clinical tools to advanced biophysical models, were compared: DTI (Diffusion Tensor Imaging), DKI (Diffusion Kurtosis Imaging), MAP-MRI (Mean Apparent Propagator), and NODDI (Neurite Orientation Dispersion and Density Imaging).

Visual comparison of structural MRI and various diffusion-weighted models (DTI, DKI, MAP-MRI, and NODDI) for a representative brain lesion.

The study population consisted of 26 males and 15 females (mean age 54.85). The HGG group included anaplastic astrocytomas and glioblastomas, while the SBM group included metastases from various primary sites. Diagnosis was confirmed via histopathology following surgery or biopsy.

Key Results and Insights

The study revealed that biophysically-informed models outperform purely mathematical ones. Specifically, the NODDI model's isotropic volume fraction ($V_{iso}$) was the most effective metric for differentiation.

  • Superior Performance of NODDI: $V_{iso}$ achieved an Area Under the Curve (AUC) of 0.871, significantly higher than metrics from other models like DKI ($RD_k$, AUC 0.760) or MAP-MRI (MSD, AUC 0.736).
  • Diagnostic Accuracy: Using an optimal threshold of 0.158, $V_{iso}$ provided a sensitivity of 95.0% and an overall accuracy of 85.4%.
  • Biological Basis: HGGs exhibited significantly lower $V_{iso}$ than SBMs. The authors suggest this is due to HGGs producing an overabundance of extracellular matrix proteins that "clutter" the interstitial space, whereas metastases tend to grow by displacing tissue and degrading the matrix, leading to higher free water isotropy.
ROC curves comparing the diagnostic performance of various metrics from five diffusion MRI models. The NODDI-Viso parameter (blue line) shows the highest area under the curve.

Interestingly, none of the models could reliably distinguish the peritumoral edema surrounding the lesions. This suggests that tumor infiltration in HGGs is too heterogeneous to be captured consistently by these global statistical measures in the swelling zone.

Limitations and Future Directions

While promising, the study is limited by its small sample size (n=41), necessitating validation in larger, multi-center cohorts. Additionally, the advanced imaging protocol required approximately 10.5 minutes of additional scan time. Future clinical adoption will likely depend on acceleration techniques, such as simultaneous multi-slice acquisition, to integrate these advanced "fingerprinting" methods into routine diagnostic workflows.

Summary updated

BibTeX
@article{mao2020differentiation,
  author = {Jiaji Mao and Weike Zeng and Qinyuan Zhang and Zehong Yang and Xu Yan and Huiting Zhang and Mengzhu Wang and Guang Yang and Min-Xiong Zhou and Jun Shen},
  title = {{Differentiation between high-grade gliomas and solitary brain metastases: a comparison of five diffusion-weighted MRI models}},
  journal = {BMC Medical Imaging},
  year = {2020},
  doi = {10.1186/s12880-020-00524-w},
}
Figure