Vascular Attributes and Malignant Brain Tumors

1
Vascular Attributes and Malignant Brain Tumors
UNC CASILab
Elizabeth Bullitt1, Guido Gerig2,3, Stephen
Aylward4, Sarang Joshi5, Keith Smith4, Matthew
Ewend1, and Weili Lin4 1Dept. of Surgery, 2Dept.
of Computer Science, 3Department of Psychiatry,
4Dept. of Radiology, and 5Department of Radiation
Oncology, University of North Carolina, Chapel
Hill, NC 27599, USA
SUMMARY
RESULTS

• We analyzed the regional morphology of blood
  vessels segmented from the MRAs of 25 brain tumor
  and16 healthy subjects.
• Parameters examined included vessel tortuosity,
  vessel density, average vessel radius, and
  terminal branch count.
• A tortuosity measure that detects tight coils and
  high-frequency sine waves successfully
  discriminated all benign from all malignant
  tumors.

METHODS
Image segmentation
Increases in ICM and TBC may occur with many
tumor types. However, we found increased
tortuosity by SOAM to be present if and only if
the tumor were malignant. Above vessels are
color-coded in relation to a malignant tumor
surface. Blueoutside, red inside, gold
traverse. Arrows point to corkscrew curves.
Example 1 Which tumor is malignant and which is
benign?
Both tumors are hypervascular with increased TBC
(red vessels) and a can of worms configuration.
The vessels at right also display corkscrew
curves. Left benign meningioma. Right
malignant metastasis.
High-resolution MR studies were obtained on 16
healthy and 25 tumor patients. Vessels were
segmented from MRA1,2. Tumors and edema were
defined from T1 and T2 images via an atlas-based
approach3.
Example 2 Which 2 tumors are malignant and which
1 is benign?
Image registration
T1GAD, T2, and MRA to be registered (rigidly)
with the same patients T1 image.
T1 image to be registered (affine) with the atlas
ICBM MRI and tissue atlas
Graph Each bar gives the number of standard
deviations from the healthy mean. All 3 tumors
are hypervascular, but abnormally high SOAM
values identify the upper and middle row tumors
as malignant (glioblastomas). The tumor in the
bottom row is infection.
CONCLUSIONS

• Analysis of vessel attributes across vessel
  populations may help diagnose and stage disease.
• Vessel density can be increased in both malignant
  and benign tumors.
• Abnormal corkscrew curves appear to be present in
  vessels of malignant tumors and absent from the
  vessels of benign tumors.
• Future work
• Validation of the current work in additional
  subjects.
• Extension to other locations such as breast or
  lung.

Vessel analyses were performed upon vessels and
vessel segments within the tumor region. Tumor
boundaries were mapped between subjects via
registration with the ICBM152 MRI atlas4, using
Rueckerts mutual information method5.
Vessel attributes

• References
• 1 Aylward S, Bullitt E (2002) Initialization,
  noise, singularities and scale in height ridge
  traversal for tubular object centerline
  extraction. IEEE-TMI 2161-75.
• 2 Bullitt E, Aylward S, Smith K, Mukherji S,
  Jiroutek M, Muller K (2001) Symbolic Description
  of Intracerebral Vessels Segmented from MRA and
  Evaluation by Comparison with X-Ray Angiograms.
  MedIA 5157-169.
• 3 Prastawa M, Bullitt E, Gerig G (2003) Robust
  estimation for brain tumor segmentation. Accepted
  MICCAI 2003.
• 4 ICBM Atlas, McConnell Brain Imaging Centre,
  Montréal Neurological Institute, McGill
  University, Montréal, Canada.
• 5 Rueckert D, Sonoda LI, Hayes C, Hill DLG,
  Leach MO, Hawkes DJ (1999) Non-rigid
  registration using free-form deformations
  Application to breast MR images. IEEE-TMI 18
  712-721
• 6 Bullitt E, Gerig G, Pizer S, Aylward SR
  (2003) Measuring tortuosity of the intracerebral
  vasculature from MRA images. IEEE-TMI
  221163-1171

• We identified 3 attributes of interest in an
  initial 5 malignant tumors
• Terminal Branch Count number of vessels within a
  tumor (TBC).
• Can of worms tortuosity, measured by Inflection
  Count Metric6 (ICM).
• Corkscrew tortuosity, measured by Sum of Angles
  metric6 (SOAM).

Images of 20 new tumors of unknown diagnosis were
then analyzed before surgery. 10 tumors were
malignant and 10 were benign.
Supported by R01 EB000219 NIH-NIBIB and R01
HL69808 NIH-HLB
MICCAI November 2003