Both Positive and Negative BOLD Signals Detected in Brain Tumor Patients

C.-S.Lin, L.M. Levy, R.C. Platenberg, K. Ward, S.S. Rajan, and D. Schellinger

Georgetown University Medical Center, Dept. of Radiology, Washington, D.C.

Introduction:

The BOLD (Blood Oxygenation Level Dependant) technique has demonstrated localized task dependant brain activation [1,2]. The studies reported have been primarily limited to analysis of positive BOLD signals. The converse oxygenation effect has not been extensively studied. We report here our observations of negative BOLD signals in brain tumor patients using a motor paradigm (complex finger tapping),

Methods:

10 normal volunteers and 9 brain tumor patients (low grade glioma) were studied. Tumors were located in the frontal and parietal lobes. Imaging was performed on a 1.5 T actively shielded whole body imaging system (Vision, Siemens, Erlangen, Germany) with a quadrature circularly polarized head coil. Both single-shot gradient-echo EPI and gradient echo pulse sequences were used to acquire images. The EPI sequence parameters: TE = 50 msec, TR = 3-4 sec, flip angle = 90, FOV = 230x230mm2, slice thickness = 4 mm, matrix size = 64x64, acquisition time per measurement = 12 sec. The functional stimulation pattern involved a 2 minute scan consisting of 25 images at each slice position with 5 off, 5 on, 5 off, 5 on, and 5 off for the EPI sequence. For the gradient echo sequences 12 images at each position with 5 off, 5 on, and 2 off were obtained. The pre and post stimulation images were compared using a trapezoid wave function. For EPI images the positive functional maps were obtained using a coefficient threshold of 0.65 which corresponds to p < 0.001. The negative functional maps were obtained by inverting the trapezoid wave function. Pixels of less than 3.5x3.5 mm2 were considered as noise and were removed from the maps. The remaining activated clusters were counted by pixel for further statistical analysis. The functional maps were then overlaid with anatomical images for better localization.

Results:

In normal controls only positive BOLD signal was observed in the motor cortex slices. The average positive signal pixel count is 320 +/- 60. Each pixel is 1.8x1.8 mm2. In the tumor patients, however, significant numbers of pixels with negative activation were detected. The patient data are summarized in the table below. Data in three patients were excluded due to misregistration artifacts.

Table:

Example of negative signal at left superior frontal cortex in normal subject. Etiologies of this phenomena if arising in brain parenchyma include subject activation during baseline; and if arising in blood vessels include inflow/outflow phenomena.

Example of negative signal at left superior frontal cortex in patient with right frontal brain tumor (grade II digodendroglioma). Causes of this phenomena are as in the normal subject; although in some cases tumor effect may be present.

Conclusions:

Normal physiological noise is about 1.0%. This applies to both positive and negative signals but can be substantially reduced by removing all the noise pixels. Negative signal is rarely seen in the normal subjects. Some visible negative signals appeared to be located in the vessels. Those signals are due to inflow artifacts. The negative signals seen in the brain tumor patients, in and near the tumor site, however, are not negligible. These signals may result from abnormal vascular physiology in and near the tumor site, and may prove of benefit in furthering neurological diagnoses. Further studies are needed to clarify what physiologic function results in the negative BOLD signal.

References:

1. Bandettini PA et al. Mag. Reson. Med. 25:2, 390 (1992).
2. Kim S-G et al. Sci. 261, 615, (1993).

Back