Intermodality or Multimodality Coregistration

Combination of functional and structural imaging provides unique information not available from independent analysis of each modality. Coregistered high-resolution anatomy in MRI or CT images provides a much more precise anatomical basis for the interpretation of functional image data. This includes improved detection of focal functional disturbances of cerebral activity, better comprehension of pathophysiologic structural-functional relationships, and more accurate identification of specific patterns of disturbances among patients or patient subpopulations of interest. Conversely, coregistered functional imaging can help interpret the clinical significance of ambiguous, subtle, or nonspecific structural lesions and other abnormalities on MRI or CT.

Most important, intermodality coregistration can aid interpretation of clinical imaging for critically important treatment decisions. Nelson et al. [10] have demonstrated the value of coregistering volumetric MRI and high-resolution [1SF]fluorodeoxyglucose PET (FDG-PET) in the evaluation of patients with brain tumors. They point out that relying on changes in MRI contrast enhancement can be misleading because of non-tumor-recurrence changes from injury to the brain caused by radiation, stroke, or benign postoperative changes. Also, active tumor may not enhance on MRI but may be metabolically active on FDG-PET. Imaging active tumor with PET scans can distinguish active tumor from radiation necrosis, but exact localization of lesion and gray matter anatomy is necessary. In the series of 24 patients reported by Nelson et al., 19 were demonstrated to benefit from MRI-PET coregistration. Particularly helped were contrast-enhancing lesions close to the cortex that distorted normal anatomy and abnormalities where absolute levels of FDG uptake were equal to or lower than that of normal gray matter.

Intermodality image coregistration is also valuable in planning radiation treatment. In the report of Rosenman et al. [11], they emphasize that the crucial step for radiation treatment planning is the accurate registration of tumor volume with simulation films (X-ray based images typically used for targeting tumor location). Otherwise, radiation targeting may include normal tissue or miss the tumor. With conventional techniques error is approximately 20%. The goal with coregistration is to transfer data from a previously acquired "non planning" study, frequently a higher resolution preintervention image, to a "planning" X-ray computed tomographic (CT) scan. Often the planning CT may not show the entire tumor, and a prior scan is needed to target the original tumor volume. This is especially a problem with lymphoma when the tumor volume is almost invariably decreased greatly from chemotherapy by the time radiation treatment planning takes place. Incomplete surgical removal of tumor is also a frequent problem where planning images do not include the original tumor volume. Coregistration of other modalities such as PET, SPECT, or magnetic resonance spectroscopy (MRS), which might identify microscopic tumor spread, part of the true tumor volume, can further enhance ability to accurately define tumor targets. This technique should be applicable in the near future with SPECT scans that are created using monoclonal antibodies binding specifically to tumor tissue proteins.

FDG-PET is a well-established epilepsy localization tool, but conventional interpretation does not typically take advantage of coregistered MRI for optimal interpretation. In a preliminary report, missed abnormalities and false interpretation of independently read-high resolution FDG-PET scans were correctly identified with coregistration to MRI, allowing an increase in both sensitivity and specificity in the detection of relative hypometabolism in patients with medial and lateral temporal lobe epilepsy [12]. It is clear that if the threshold for detection of subtle abnormalities on FDG-PET is to be lowered, coregistration to MRI is necessary to correct anatomic asymmetries, to identify whether any structural anomaly may be the cause of a focal defect, and to determine if a suspected region is affected by partial volume averaging. Figure 5 shows an FDG-PET scan from a patient with left mesial temporal lobe epilepsy. The scan was interpreted as normal or nonlateralizing. Only with coregistered MRI anatomy and partial volume average correction was unequivocal focal hypometabolism in the left hippocampus identified.

Localization of the epileptogenic zone, the focal region of brain responsible for generation of seizures, is the goal of preoperative epilepsy surgery evaluations. This process utilizes extensive imaging and clinical neurophysiolgic localization techniques in what is considered a multimodality examination of the brain. Electroencephalographic (EEG) or magneto-encephalographic (MEG) source localization can be combined with both functional and structural anatomic imaging in an attempt to identify the epileptogenic zone. This is especially helpful in cases that do not have an identifiable epileptogenic lesion on MRI [13], for successful surgery is much less likely in cases where an epileptogenic substrate has not been confidently

FIGURE 5 MRI to FDG-PET coregistration — volume of interest (VOI) analysis with partial volume correction — image analysis from a 45-year-old woman with left mesial temporal lobe epilepsy. Top left MRI images show hippocampal VOIs manually drawn on precise coronal MRI anatomy. Middle images show VOI contours resliced in the original FDG-PET sinogram space. Bottom right images show the convolved MRI gray matter coregistered to PET space; these images were used to correct for partial volume averaging. FDG-PET images were interpreted as normal (no relative lateralized or focal hypometabolism). Before partial volume correction, hippocampal VOI specific activity of FDG was symmetric (normal or nonlateralizing). After partial volume average correction, an 18% asymmetry of hippocampal metabolism was demonstrated in the left atrophic hippocampus. See also Plate 98.

FIGURE 5 MRI to FDG-PET coregistration — volume of interest (VOI) analysis with partial volume correction — image analysis from a 45-year-old woman with left mesial temporal lobe epilepsy. Top left MRI images show hippocampal VOIs manually drawn on precise coronal MRI anatomy. Middle images show VOI contours resliced in the original FDG-PET sinogram space. Bottom right images show the convolved MRI gray matter coregistered to PET space; these images were used to correct for partial volume averaging. FDG-PET images were interpreted as normal (no relative lateralized or focal hypometabolism). Before partial volume correction, hippocampal VOI specific activity of FDG was symmetric (normal or nonlateralizing). After partial volume average correction, an 18% asymmetry of hippocampal metabolism was demonstrated in the left atrophic hippocampus. See also Plate 98.

identified prior to surgery. Sometimes the pathology is cryptogenic —not visible on conventional MRI. In other cases an abnormality on MRI is nonspecific or of uncertain clinical significance. In yet others a MRI abnormality is ambiguous (e.g., large cystic lesions) or even misleading (e.g., mesial temporal sclerosis in patients with lateral temporal lobe epilepsy— dual pathology). In all of these circumstances multimodality coregistration with functional imaging and electrophysiologic localization can improve the accuracy of identifying the epileptogenic zone. Figure 6 illustrates a true example of cryptogenic pathology that was detected only with multimodality coregistration. In this case the patient had seizures suspected to arise from the right temporal lobe based on clinical features and frequent interictal epileptiform discharges (spikes) recorded on scalp EEG. MRI showed no lesion or abnormality. An FDG-PET scan was originally interpreted as normal. Intracranial electrode recording of seizures did not localize or even confidently define the hemisphere from which the seizures arose. MEG estimated the 3D location of epileptiform abnormalities to a focal region of the right lateral temporal lobe. Reexamination of the coregistered FDG-PET demonstrated unequivocal relative

FIGURE 6 MEG to MRI to FDG-PET coregistration — discovery of cryptogenic epileptogenic pathology-image analysis from a 28-year-old woman with right temporal lobe epilepsy. (a) FDG-PET overlaid on MRI shows questionable relative focal hypometabolism in the right superior temporal gyrus. (b, left) MEG spike dipole source localization (white triangles, tails represent dipole orientation) overlaid on a selected coregistered coronal MRI slice from the posterior temporal lobes — dipole sources correspond to an active epileptiform disturbance of cerebral activity with scalp EEG (epileptiform sharp waves maximum at right mid-temporal electrode); MRI is normal. (Right) Corresponding FDG-PET slice — focal hypometabolism present in the right superior temporal gyrus (large white arrow) is remarkably colocalized with MEG spike sources. Histopathology revealed cryptogenic hamartomatous dysplasia. See also Plate 99.

FIGURE 6 MEG to MRI to FDG-PET coregistration — discovery of cryptogenic epileptogenic pathology-image analysis from a 28-year-old woman with right temporal lobe epilepsy. (a) FDG-PET overlaid on MRI shows questionable relative focal hypometabolism in the right superior temporal gyrus. (b, left) MEG spike dipole source localization (white triangles, tails represent dipole orientation) overlaid on a selected coregistered coronal MRI slice from the posterior temporal lobes — dipole sources correspond to an active epileptiform disturbance of cerebral activity with scalp EEG (epileptiform sharp waves maximum at right mid-temporal electrode); MRI is normal. (Right) Corresponding FDG-PET slice — focal hypometabolism present in the right superior temporal gyrus (large white arrow) is remarkably colocalized with MEG spike sources. Histopathology revealed cryptogenic hamartomatous dysplasia. See also Plate 99.

focal hypometabolism colocalized to the MEG estimates of epileptiform abnormalities. Resective surgery in this region eliminated the patient's seizures, and histopathology (microscopic examination of the surgically resected tissue) revealed focal cortical dysplasia, a classic pathology frequently associated with seizures, but also one that may or may not be detected with high-quality MRI.

When further combined with fMRI or PET-based brain mapping, a completely noninvasive presurgical epilepsy evaluation can be envisioned. Before the presurgical evaluation can become completely noninvasive, rigorous validation must be accomplished with new methods measured against standard intracranial localization techniques. Multimodality coregistra-tion is already widely used in laboratories performing brain mapping research with PET, fMRI, EEG, and MEG. Combining neurophysiologic and neuroimaging modalities enables combination and validation of source localization techniques, basic research that is required to advance the application of source localization in epilepsy [14]. Techniques for modeling dipole sources should theoretically benefit from including anatomical constraints about volume conduction and orientation of current flow. A priori knowledge about function localization from relatively high spatial resolution imaging modalities (fMRI and PET) should help exploit source localization with high temporal resolution modalities (EEG and MEG). These advances are being worked out with functional mapping paradigms but will ultimately benefit epilepsy localization, including interictal/ictal neurophysiology, metabolic function, and detection of seizure propagation patterns.

Electrode to image registration is necessary in order to perform validation of neurophysiologic localization. With EEG both scalp and intracranial electrode positions can be co-registered to MRI. The patient's scalp surface and electrode positions can be mapped with a 3D magnetic digitizer [15]. Surface contours, landmarks [16], or fiducial markers can be used for coregistration. An additional approach is to physically measure the coordinates of EEG electrodes and align them with the scalp surface on an MRI [17,18]. Intracranial EEG electrode positions can be coregistered to other imaging modalities by simply imaging the patient with the electrodes implanted. Imaging of electrodes for multimodality registration purposes is best performed with MRI or CT, but skull radiographs can also be used [19]. If MRI is performed, special (and more expensive) electrodes made of nonferromagnetic nickel chromium alloys should be used to eliminate risk of electrode movement within the patient's head. Whether CT or MRI is used, artifact obscures the exact visualization of the contacts (10 mm signal void on MRI). One option to eliminate artifact and reduce cost is to coregister a postimplantation CT to a preimplantation MRI (Fig. 7).

The surgical pathology correlates of many examples of MRI signal alterations, especially isolated focal increased T2 signal and relative focal hypometabolism on FDG-PET, remain unclear. In patients with epilepsy and an isolated finding of focal increased T2 signal, the temptation is to attribute the change to a pathologic diagnosis of gliosis or cortical dysplasia; but in a report by Mitchell et al. [20], such changes were not associated with any histopathologic abnormality. The surgical specimens, however, were not imaged and coregistered to presurgical scans. In one of the first reports of mapping histology to metabolism using coregistration, stained whole brain sections from Alzheimer's disease patients were optically imaged, digitally reconstructed into a 3D volume, and coregistered to premortem FDG-PET scans [21]. Areas of decreased FDG uptake did not correlate with neural fibrillary tangle staining density, a commonly presumed relationship before this study. Although some technical difficulties remain for handling the imaging of surgical specimens, this area of coregistration holds great promise for accurately understanding the underlying pathology of image signal changes, an aspect of diagnostic radiology that has become increasingly difficult with rapid ongoing advances in neuroimaging techniques.

0 0

Responses

  • robert
    How is coregistration of spect and ct data performed?
    5 years ago

Post a comment