F. Reed Murtagh

Automatic segmentation of non-enhancing brain tumors in magnetic resonance images

Tumor segmentation from magnetic resonance (MR) images may aid in tumor treatment by tracking the progress of tumor growth and/or shrinkage. In this paper we present the first automatic segmentation method which separates non-enhancing brain tumors from healthy tissues in MR images to aid in the task of tracking tumor size over time. The MR feature images used for the segmentation consist of three weighted images (T1, T2 and proton density (PD)) for each axial slice through the head. An initial segmentation is computed using an unsupervised fuzzy clustering algorithm. Then, integrated domain knowledge and image processing techniques contribute to the final tumor segmentation. They are applied under the control of a knowledge-based system. The system knowledge was acquired by training on two patient volumes (14 images). Testing has shown successful tumor segmentations on four patient volumes (31 images). Our results show that we detected all six non-enhancing brain tumors, located tumor tissue in 35 of the 36 ground truth (radiologist labeled) slices containing tumor and successfully separated tumor regions from physically connected CSF regions in all the nine slices. Quantitative measurements are promising as correspondence ratios between ground truth and segmented tumor regions ranged between 0.368 and 0.871 per volume, with percent match ranging between 0.530 and 0.909 per volume.

Diffusion tensor imaging tractography in patients with intramedullary tumors: comparison with intraoperative findings and value for prediction of tumor resectability

OBJECT The aim of this retrospective study was to evaluate the predictive value of diffusion tensor (DT) imaging with respect to resectability of intramedullary spinal cord tumors and to determine the concordance of this method with intraoperative surgical findings. METHODS Diffusion tensor imaging was performed in 14 patients with intramedullary lesions of the spinal cord at different levels using a 3-T magnet. Routine MR imaging scans were also obtained, including unenhanced and enhanced T1-weighted images and T2-weighted images. Patients were classified according to the fiber course with respect to the lesion and their lesions were rated as resectable or nonresectable. These results were compared with the surgical findings (existence vs absence of cleavage plane). The interrater reliability was calculated using the kappa coefficient of Cohen. RESULTS Of the 14 patients (7 male, 7 female; mean age 49.2 +/- 15.5 years), 13 had tumors (8 ependymomas, 2 lymphomas, and 3 astrocytoma). One lesion was proven to be a multiple sclerosis plaque during further diagnostic workup. The lesions could be classified into 3 types according to the fiber course. In Type 1 (5 cases) fibers did not pass through the solid lesion. In Type 2 (3 cases) some fibers crossed the lesion, but most of the lesion volume did not contain fibers. In Type 3 (6 cases) the fibers were completely encased by tumor. Based on these results, 6 tumors were considered resectable, 7 were not. During surgery, 7 tumors showed a good cleavage plane, 6 did not. The interrater reliability (Cohen kappa) was calculated as 0.83 (p < 0.003), which is considered to represent substantial agreement. The mean duration of follow-up was 12.0 +/- 2.9. The median McCormick grade at the end of follow-up was II. CONCLUSIONS These preliminary data suggest that DT imaging in patients with spinal cord tumors is capable of predicting the resectability of the lesion. A further prospective study is needed to confirm these results and any effect on patient outcome.

MRI evidence of mesial temporal sclerosis in patients with psychogenic nonepileptic seizures

Mesial temporal sclerosis (MTS) is the most common cause of temporal lobe epilepsy, and MRI evidence of MTS is critical in the evaluation for temporal lobectomy.1,2⇓ For those reasons, MRI is now commonly performed specifically for the purpose of identifying MTS. However, the frequency of significant MRI evidence of MTS in patients without seizure (i.e., specificity) is unknown. We report four cases that illustrate an increasingly important pitfall in the evaluation of patients with seizures. Patients were 28 to 35 years of age and had onset of “seizures” between 25 and 30 years of age (duration of seizures 3 to 6 years). They were all referred for possible epilepsy surgery after an MRI showed evidence of MTS. MRI (1.5 tesla) used 1-mm sections and was performed using fast spin echo T2 (repetition time [TR] 5500, echo time [TE] 21.3, 256 × 256 matrix) coronal sequence, inversion recovery sequence (three-dimensional/fast-spin echo gradient recall = 30, TR 10.2, TE 2), axial T2 images at a …

MRI of the brain is safe in patients implanted with the vagus nerve stimulator

Metallic devices generally represent a contra-indication for MRI scanning. Based on laboratory testing, the neuro cybernetic prosthesis (NCP) is labelled MRI compatible when used with a send and receive head coil. However, there are no published clinical data to support the safety of brain MRI in patients with the NCP. Our objective was to report clinical experience with such a population. We questioned 40 centres that had implanted the NCP system as of 10/1/99. If MRI had been performed on any vagus nerve stimulator patients, we collected information on these patients, the MRI technique used, any events noted during the scan, including both subjective reports (by the patient ), and observable (objective) changes noted by the staff. Twelve centres (30%) responded. Over a time period of 3 years, there were a total of 27 MRI scans performed in 25 patients. All scanners were 1.5 T. A head coil was used in 26 scans, and a body coil in one. The indications for the scans were diverse. Seven were related to the epilepsy, including aetiology or pre-surgical evaluation. Others were unrelated, including brain tumours, cerebral haematoma, vasculitis, headaches, and head trauma. Three scans were performed with the stimulator on, while 24 were performed with the stimulator off. One patient had a mild objective voice change for several minutes. No other objective changes were noted in any of the patients. One 11-year old reported chest pain while experiencing severe claustrophobia. Twenty-five patients denied any discomfort around the lead or the generator. We conclude that this clinical series supports the safety of routine brain MRI using a send and receive head coil in patients implanted with the NCP System.

Use of a clinical MR scanner for imaging the rat brain

Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy are established techniques that enable noninvasive anatomic and functional tissue characterization in vivo. These tools have been employed to probe experimental models of neoplasia, cerebrovascular disease, brain injury, and neurotransplantation in small animals. To date, these studies have been executed primarily on research-dedicated instruments of limited availability or resolution. Using relatively straightforward software and hardware modifications of a widely used clinical MRI unit, we were able to image numerous structures within the living rat brain including the neostriatum, hippocampus, periaqueductal gray, and the ventricular system. Illustrative applications of this imaging technique in two intracerebral infusion models involving rats are presented. Such adaptation of clinical MRI scanners has the potential to significantly expand the availability of high resolution in vivo imaging of small animals for a variety of experimental protocols.

Computer tomographically guided discography as a determinant of normal disc level before fusion

Computer tomographically guided discography was used as a reproducible reliable method of determining the internal architecture and integrity of intervertebral discs before spinal fusion operation. One hundred disc levels were studied in 60 prospective patients with 54 [54%) of levels proving abnormal. Twenty-nine levels (29% of total levels) demonstrated complete anular tears, and 25 (25% of total) showed incomplete radial tears of the anulus, often multiple, indicative of degenerative change. In each of these cases, one level higher had to be studied and was subsequently found to be normal and would, therefore, support a fusion.Computer tomographically guided discography was used as a reproducible and reliable method of determining the internal architecture and integrity of intervertebral discs before spinal fusion operation. One hundred disc levels were studied in 60 prospective patients with 54 (54%) of levels proving abnormal. Twenty-nine levels (29% of total levels) demonstrated complete anular tears, and 25 (25% of total) showed incomplete radial tears of the anulus, often multiple, indicative of degenerative change. In each of these cases, one level higher had to be studied and was subsequently found to be normal and would, therefore, support a fusion.

Magnetic resonance imaging of intracranial epidermoid tumours

Epidermoid tumours are rare lesions within the skull which are usually found in a cerebellopontine angle. In the computed tomography (CT) literature, they have been described as well-defined lesions of low density which seldom show calcification or tissue enhancement. We have recently encountered three intracranial epidermoids, which appeared to exhibit uniform findings on low field strength magnetic resonance (MR) of moderate signal intensity in the T1-weighted images and a bright signal (increased T2) in the T2-weighted sequences. It may be possible to predict accurately the presence of an epidermoid tumour when this particular configuration of findings is present.

CT localization of posterior fossa bleeding sites in subarachnoid hemorrhage

Abstract Of 155 consecutive patients who were initially seen with symptoms of subarachnoid hemorrhage, 5 (3.2%) were ultimately diagnosed as having aneurysms or arteriovenous malformations located beneath the tentorium within the posterior fossa. Three aneurysms involved the vertebral arteries, and one aneurysm involved the superior cerebellar artery; the single arteriovenous malformation had feeding vessels from the superior cerebellar artery as well as meningeal vessels. Three of these 5 patients showed no localizing CT characteristics such as blood in the sub-arachnoid spaces on the initial scan. Four showed evidence of brainstem swelling, possibly secondary to vasospasm, and 4 showed intraventricular obstructive hydrocephalus. The sites of the bleeding appear to be more difficult to diagnose with computed tomography than supratentorial lesions and require a high degree of suspicion for diagnosis to be made with angiography.

MRI of spinal cord radiation necrosis simulating recurrent cervical cord astrocytoma and syringomyelia

BACKGROUND Although primary intramedullary tumors of the spinal cord with syrinx formation are well documented, there have been no reports of extensive syrinx formation or cystic degeneration associated with radiation necrosis. METHODS We report a case of radiation necrosis and syrinx formation in a 49-year-old woman with a 5-year history of astrocytoma grade II of the cervical cord, who progressed to quadriparesis following surgery and radiation therapy. Magnetic resonance imaging (MRI) of the cervical and thoracic spine demonstrated enlargement of upper cervical cord (C1-C6) with diffuse increased signal enhancing mass by gadolinium, as well as appearance of syrinx from T4-T10. RESULTS Autopsy findings indeed revealed a small, residual, infiltrating glioma in the upper cervical areas, but the diffuse parenchymal abnormality seen on MRI as prolonged T2 characteristics on double-echo spin-echo sequence was revealed to be radiation necrosis. CONCLUSION What appeared to be a cystic cavity or syrinx at the thoracic level was also diagnosed as radiation necrosis with cyst formation on histologic examination.

AUTOMATED QUANTIFICATION OF BRAIN MAGNETIC RESONANCE IMAGE HYPERINTENSITIES USING HYBRID CLUSTERING AND KNOWLEDGE-BASED METHODS

Previous computerized methods of hyperintensity identification in brain magnetic resonance images (MRI) either rely heavily on human intervention or on simple thresholding techniques. Such methods can lead to considerable variation in the quantification of brain hyperintensities depending upon image parameters such as contrast. This paper describes an automated, knowledge‐guided method of hyperintensity detection in brain MRI that addresses problems associated with human subjectivity and thresholding techniques. This method, which we call knowledge‐guided hyperintensity detection (KGHID), uses encoded knowledge of brain anatomy and MRI characteristics of individual tissues to reclassify pixels from an initial unsupervised segmentation. With this encoded knowledge, KGHID discriminates lesions embedded within the white matter, hyperintense lesions of the basal ganglia and the periventricular ring. The method is designed for high sensitivity detection and monitoring of subtle lesions in patients with neurodegenerative diseases.