Article
First step to a probabilistic model determining the origin of brain metastases – A MRI study
Search Medline for
Authors
Published: | June 18, 2018 |
---|
Outline
Text
Objective: Fifteen percent of brain metastases (BM) are due to an unknown primary tumor. We evaluated the MRI characteristics of BM with multiple imaging protocols to calculate a probability score of the tumor origin.
Methods: MR images were analyzed based on 35 previously determined criteria, such as number, size and location of BM, edema volume, as well as homogeneity, intensity and character of contrast enhancement. Analyses were performed blinded to primary tumor. Comparison of MRI characteristics was done using χ2 test.
Results: More than 458 BM of 189 patients were evaluated. BM originating from the gastrointestinal tract (GIT) were more than coincidently less multiple (p=0.051), more often had a polylobular form (p=0.03) and seemed to be more inhomogeneous (p=0.04) in T1-weighted images than BM from the lung. BM from the GIT had a median size of 39,13cm3 and on average were larger than BM from other primary sites, though not statistically significant (p=0.36). BM of malignant melanoma on the other hand were smaller than other BM (p=0.069) with an average size of 17.7cm3 and were more than coincidently often homogenously contrast enhancing (p<0.001) than other BM. BM from breast (p=0.02) and the GIT (p=0.02) were less likely to occur in the frontal lobe than tumors originating from the lungs. BM from breast are more likely to have meningeal contact (p=0.04) and meningeal enhancement (p=0.007) then all other entities. BM of the breast (p=0.006), skin (p=0.02) and gastrointestinal tract (p=0.03) were more likely to have central necrosis than others.
Conclusion: Our study reveals that a probabilistic model to determine the origin of BM is possible by grouping BM of different origin by MRI appearance. To create a score with high power that can be used in clinical routine, the number of BM to be analyzed has to be tripled.