Exploiting Graphoelements and Convolutional Neural Networks with Long Short Term Memory for Classification of the Human Electroencephalogram.

The electroencephalogram (EEG) is a cornerstone of neurophysiological research and clinical neurology. Historically, the classification of EEG as showing normal physiological or abnormal pathological activity has been performed by expert visual review. The potential value of unbiased, automated EEG classification has long been recognized, and in recent years the application of machine learning methods has received significant attention. A variety of solutions using convolutional neural networks (CNN) for EEG classification have emerged with impressive results. However, interpretation of CNN results and their connection with underlying basic electrophysiology has been unclear. This paper proposes a CNN architecture, which enables interpretation of intracranial EEG (iEEG) transients driving classification of brain activity as normal, pathological or artifactual. The goal is accomplished using CNN with long short-term memory (LSTM). We show that the method allows the visualization of iEEG graphoelements with the highest contribution to the final classification result using a classification heatmap and thus enables review of the raw iEEG data and interpret the decision of the model by electrophysiology means.

Ictal SPECT statistical parametric mapping in temporal lobe epilepsy surgery.

OBJECTIVE: Although subtraction ictal SPECT coregistered to MRI (SISCOM) is clinically useful in epilepsy surgery evaluation, it does not determine whether the ictal-interictal subtraction difference is statistically different from the expected random variation between 2 SPECT studies. We developed a statistical parametric mapping and MRI voxel-based method of analyzing ictal-interictal SPECT difference data (statistical ictal SPECT coregistered to MRI [STATISCOM]) and compared it with SISCOM. METHODS: Two serial SPECT studies were performed in 11 healthy volunteers without epilepsy (control subjects) to measure random variation between serial studies from individuals. STATISCOM and SISCOM images from 87 consecutive patients who had ictal SPECT studies and subsequent temporal lobectomy were assessed by reviewers blinded to clinical data and outcome. RESULTS: Interobserver agreement between blinded reviewers was higher for STATISCOM images than for SISCOM images (kappa = 0.81 vs kappa = 0.36). STATISCOM identified a hyperperfusion focus in 84% of patients, SISCOM in 66% (p < 0.05). STATISCOM correctly localized the temporal lobe epilepsy (TLE) subtypes (mesial vs lateral neocortical) in 68% of patients compared with 24% by SISCOM (p = 0.02); subgroup analysis of patients without lesions (as determined by MRI) showed superiority of STATISCOM (80% vs 47%; p = 0.04). Moreover, the probability of seizure-free outcome was higher when STATISCOM correctly localized the TLE subtype than when it was indeterminate (81% vs 53%; p = 0.03). CONCLUSION: Statistical ictal SPECT coregistered to MRI (STATISCOM) was superior to subtraction ictal SPECT coregistered to MRI for seizure localization before temporal lobe epilepsy (TLE) surgery. STATISCOM localization to the correct TLE subtype was prognostically important for postsurgical seizure freedom.

Effects of 5-HT(3) antagonism on postprandial gastric volume and symptoms in humans.

BACKGROUND: Alosetron reduces symptoms of dyspepsia, but the physiological basis for the symptomatic benefit is unclear. AIM: To assess 5-HT3 antagonism on postprandial gastric volume and symptoms after ingestion of maximum tolerable volume of a liquid meal. METHODS: In 36 healthy volunteers, we assessed effects of placebo, 0.5 and 1 mg b.d. alosetron on fasting and postprandial gastric volumes (using single photon emission computed tomography) and symptoms based on 100 mm VAS, 30 min after maximum volume ingested. RESULTS: The 5-HT3 antagonist reduced postprandial symptoms (aggregate score: P < 0.05), nausea (P < 0.001), and tended to reduce bloating (P=0.08). Both 0.5 and 1 mg alosetron reduced nausea (P < 0.025); 1 mg alosetron reduced aggregate symptoms (P < 0.05) and bloating (P < 0.05). Effects on pain (P=0.19) and fullness (P=0.14) were not statistically significant. There were no significant effects of the 5-HT3 antagonist on volume of meal tolerated or on SPECT-measured fasting or postprandial gastric volumes. CONCLUSION: 5-HT3 antagonism reduces aggregate symptoms, nausea and bloating after a liquid meal without increase in gastric volumes, suggesting a role for 5-HT3 in afferent functions in healthy humans during the postprandial period.

Noninvasive measurement of gastric accommodation in patients with idiopathic nonulcer dyspepsia.

OBJECTIVES: Postprandial symptoms are associated with impaired postprandial gastric accommodation. The aims of this study were to apply a noninvasive method to measure accommodation of the entire stomach in healthy subjects and in patients with idiopathic dyspeptic symptoms, and to assess the frequency of abnormal gastric accommodation and emptying of solids in these patients. METHODS: In 20 healthy volunteers and 32 tertiary referral patients, we used i.v. 99mTc-single photon emission computed tomography (SPECT) to measure fasting and postprandial gastric volumes; we expressed the volume response to feeding ("accommodation") as the change in gastric volume and the ratio of postprandial/fasting volumes. The stomach was identified in transaxial SPECT tomographic images using a semiautomated, intensity-based extraction algorithm. Whole gastric volumes were measured using AnalyzeAVW software. Gastric emptying in patients was measured by scintigraphy. We also assessed dyspeptic symptoms and the association with normal or reduced accommodation. RESULTS: SPECT imaging detects the postprandial change in gastric volume ("accommodation") in health and disease. Among healthy subjects (eight men, 12 women), the postprandial/fasting gastric volume ratio was 4.9+/-1.7 (mean +/- SD; fifth through 95th percentiles 3-8, median 4.6). Thirteen (41%) patients with idiopathic nonulcer dyspepsia had reduced postprandial "accommodation." Gastric emptying was fast in four (13%), normal in 25 (78%), and slow in three (9%) patients. Both tests were normal in 50% of patients. Weight loss of >10 pounds tended to be more frequently observed in those with reduced "accommodation" (62% vs 32%, p = 0.09). CONCLUSIONS: SPECT imaging noninvasively measures fasting and postprandial gastric volumes in humans. Half the patients with idiopathic nonulcer dyspepsia had impaired gastric accommodation or emptying. Reduced gastric "accommodation" was observed in 41% of a group with idiopathic nonulcer dyspepsia. Abnormal gastric emptying is less frequent (22%).

Subtraction peri-ictal SPECT is predictive of extratemporal epilepsy surgery outcome.

OBJECTIVES: To determine whether localization of extratemporal epilepsy with subtraction ictal SPECT coregistered with MRI (SISCOM) is predictive of outcome after resective epilepsy surgery, whether SISCOM images provide prognostically important information compared with standard tests, and whether blood flow change on SISCOM images is useful in determining site and extent of excision required. BACKGROUND: The value of SISCOM in predicting surgical outcome for extratemporal epilepsy is unknown, especially if MRI findings are nonlocalizing. METHODS: SISCOM images in 36 consecutive patients were classified by blinded reviewers as "localizing and concordant with site of surgery," "localizing but nonconcordant with site of surgery," or "nonlocalizing." SISCOM images were coregistered with postoperative MRI, and reviewers visually determined whether cerebral cortex underlying the SISCOM focus had been completely resected, partially resected, or not resected. RESULTS: Twenty-four patients (66.7%) had localizing SISCOM, including 13 (76.5%) of those without a focal MRI lesion. Eleven of 19 patients (57.9%) with localizing SISCOM concordant with the surgical site, compared with 3 of 17 (17.6%) with nonlocalizing or nonconcordant SISCOM, had an excellent outcome (p < 0.05). With logistic regression analysis, SISCOM findings were predictive of postsurgical outcome, independently of MRI or scalp ictal EEG findings (p < 0.05). The extent of resection of the cortical region of the SISCOM focus was significantly associated with the rate of excellent outcome (100% with complete resection, 60% with partial resection, and 20% with nonresection, p < 0.05). CONCLUSION: SISCOM images may be useful in guiding the location and extent of resection in extratemporal epilepsy surgery.

Voxel significance mapping using local image variances in subtraction ictal SPET.

Subtraction ictal SPET co-registered to MRI (SISCOM) has been shown to aid epileptogenic localization and improve surgical outcome in partial epilepsy patients. This paper reports a method of identifying significant areas of epileptogenic activation in the SISCOM subtraction image, taking into account normal variation between sequential 99Tcm-ethyl cysteinate diethylester SPET scans of single individuals, and attempts to assess the clinical value of statistical mapping in subtraction SPET. Non-linear inter-subject registration is used to combine a group of subtraction images into a common anatomical framework. A map of the pixel intensity standard deviation values in the subtraction images is created, and this map is non-linearly registered to a patient's SISCOM subtraction image. Pixels in the patient subtraction image were then evaluated based upon the statistical characteristics of corresponding pixels in the atlas. SISCOM images created with the voxel variance method were rated higher in quality than the conventional image variance method in 15 patients. No difference in localization rate was observed between the voxel variance mapping and image variance methods. The voxel significance mapping method was shown to improve the quality of clinical SISCOM images.

Localization of the epileptic focus by low-resolution electromagnetic tomography in patients with a lesion demonstrated by MRI.

Patients with medically intractable partial epilepsy and well-defined symptomatic MRI lesions were studied using phase-encoded frequency spectral analysis (PEFSA) combined with low-resolution electromagnetic tomography (LORETA). Ten patients admitted to the epilepsy monitoring unit with MRI-identified lesions and intractable partial epilepsy were studied using 31-electrode scalp EEG. The scalp electrodes were located in three-dimensional space using a magnetic digitizer and coregistered with the patient's MRI. PEFSA was used to obtain a phase-encoded scalp map for the ictal frequencies. The ictal generators were obtained from the scalp map using LORETA. In addition, the generators of interictal epileptogenic spikes were identified using time-domain LORETA. The LORETA generators were rostral to the MRI lesion in 87% (7/8) of patients with temporal lobe lesions, but all were located in the mesial temporal lobe in concordance with the patients' MRI lesions. In patients with frontal lobe epilepsy, the ictal generators at the time that the spectral power was maximal localized to the MRI lesions. Eight of 10 patients had interictal spikes, of which 4 were bilateral independent temporal lobe spikes. Only generators of the interictal spikes that were ipsilateral to seizure onset correlated with the ictal generators. LORETA combined with PEFSA of the ictal discharge can localize ictal EEG discharges accurately and improve correlation with brain anatomy by allowing coregistration of the ictal generator with the MRI. Analysis of interictal spikes was less useful than analysis of the ictal discharge.

Subtraction ictal SPECT coregistered to MRI for seizure focus localization in partial epilepsy.

Peri-ictal single-photon emission computed tomography (SPECT) of the brain is increasingly used in localizing the seizure focus in presurgical evaluation of patients with partial epilepsy. However, traditional side-by-side visual interpretation of ictal and interictal SPECT films is hampered by differences in slice location and tracer activity. Precise correlation of the seizure focus with a high-quality image of the underlying brain anatomy can improve the physician's understanding of seizure neurophysiology and assist in surgical planning. Computer-based methods have been developed for aligning, normalizing, and subtracting digital ictal and interictal SPECT images of the patient's brain to produce a map of the blood flow changes occurring between the seizure and resting states. These maps are then aligned with a high-resolution magnetic resonance image (MRI) of the patient's brain anatomy and fused to identify anatomical regions involved in the seizure. The purpose of this article is to review the technical components and clinical implementation of subtraction ictal SPECT, as well as to discuss recent technological advances that could extend and improve the diagnostic and localizing capacity of this method.

The EEG evaluation of single photon emission computed tomography abnormalities in epilepsy.

Single photon emission computed tomography (SPECT) has increasingly been used as a diagnostic procedure for localizing epileptic seizure foci and as a research tool for investigating the physiologic mechanisms underlying seizure activity. With increasing use of SPECT in localizing the seizure focus for epilepsy surgery, there arises a need to critically assess its current role in the evaluation of patients for epilepsy surgery, especially as it relates to other clinical and laboratory data used in presurgical evaluation. Ictal EEG discharge has traditionally been used as the "gold standard" against which SPECT studies are compared in assessing the latter's localizing value. However, this practice presents a major challenge because SPECT studies are often reserved for patients with nonlocalizing EEG or magnetic resonance imaging findings. Nonetheless, SPECT studies in evaluation for epilepsy surgery should always be performed with the knowledge of the patient's EEG activity preceding, during, and after the injection of the radiotracer. The advent of techniques such as subtraction SPECT with co-registration on magnetic resonance imaging (SISCOM) and computer image-guided surgery has great potential in enhancing the clinical electrophysiologic evaluation of SPECT-detected abnormalities in epilepsy. These techniques permit accurate spatial correlation between intracranial EEG activity and SPECT perfusion patterns. The techniques can also be used to evaluate the effect of the extent of EEG focus resection compared with that of SISCOM focus resection to determine which has more prognostic importance in postsurgical control of seizures. Both animal and human studies are warranted to advance our knowledge of the electrophysiology associated with the various SPECT perfusion patterns.

Development of a test to measure gastric accommodation in humans.

Postprandial symptoms of bloating, distension, early satiety, and nausea are associated with impaired postprandial gastric accommodation, which is detectable by means of an intragastric, barostatically controlled balloon in the proximal stomach and by ultrasound in the distal stomach. Our aim was to develop a noninvasive method to measure the entire gastric accommodation reflex. In 10 healthy volunteers, we used single photon emission computed tomography (SPECT) to measure fasting and postprandial gastric volumes. This method involved intravenous injection of (99m)Tc pertechnetate and gastric reconstruction of tomographic images with Analyze software. SPECT-Analyze imaging detects the postprandial gastric accommodation reflex in vivo. Mean fasting gastric volume was 182 +/- 11 (SE) ml and mean postprandial volume was 690 +/- 32 ml (P < 0.001). Both proximal and distal segments of stomach showed a two- to almost fourfold difference in volumes postprandially. Intraobserver coefficients of variation in estimated fasting and postprandial volumes were 9 and 8%; interobserver variations were 13 and 12%, respectively. SPECT-Analyze noninvasively measures postprandial gastric (total, proximal, and distal) accommodation in humans. This method appears promising to compare the accommodation response in health and disease and to perform mechanistic studies of the accommodation response.

Quantitative and clinical analysis of SPECT image registration for epilepsy studies.

UNLABELLED: This study reports quantitative measurements of the accuracy of two popular voxel-based registration algorithms--Woods' automated image registration algorithm and mutual information correlation--and compares these with conventional surface matching (SM) registration. METHODS: The registration algorithms were compared (15 different matches each) for (a) three-dimensional brain phantom images, (b) an ictal SPECT image from a patient with partial epilepsy matched to itself after modification to simulate changes in the cerebral blood flow pattern and (c) ictal/interictal SPECT images from 15 patients with partial epilepsy. Blinded visual ranking and localization of the subtraction images derived from the patient images were also performed. RESULTS: Both voxel-based registration methods were more accurate than SM registration (P < 0.0005). Automated image registration algorithm was more accurate than mutual information correlation for the computer-simulated ictal/interictal images and the patient ictal/interictal studies (P < 0.05). The subtraction SPECTs from SM were poorer in visual ranking more often than the voxel-based methods (P < 0.05). CONCLUSION: Voxel intensity-based registration algorithms provide significant improvement in ictal/interictal SPECT registration accuracy and result in a clinically detectable improvement in the subtraction SPECT images.

Dual-isotope SPECT using simultaneous acquisition of 99mTc and 123I radioisotopes: a double-injection technique for peri-ictal functional neuroimaging.

UNLABELLED: The acquisition of multiple radiotracer studies at different time points during a neurological event permits the study of different functional activation states in humans. Peri-ictal SPECT is a promising technique for localizing the epileptogenic zone and would be enhanced by the ability to acquire sequentially coregistered ictal and postictal SPECT images of a single seizure. This study was designed to develop and validate an accurate method for the simultaneous acquisition of 99mTc and 123I SPECT images of the brain. METHODS: A multicompartment, transaxial Hoffman brain-slice phantom was filled with 99mTc, 123I or a 3:1 mixture of the two isotopes. Planar and SPECT images were acquired by a dual-head gamma camera system equipped with parallel and fanbeam collimators, respectively. Thirty-two energy windows (2 keV width) were acquired over the energy range 120-184 keV. From the planar data, the signal-to-noise characteristics and crosstalk were measured for each energy window and used to devise an energy window acquisition strategy that was then applied to the SPECT data. Three summed energy windows were created: a primary 99mTc image (130-146 keV), a primary 123I image (152-168 keV) and a secondary 99mTc crosstalk image (134-140 keV). A fraction (0.041) of the 99mTc crosstalk image was subtracted from the 123I image. No crosstalk correction was performed on the primary 99mTc image. RESULTS: (a) Planar images: results showed 1.3% crosstalk in the 123I image compared with 19.7% for a 10% asymmetric energy window alone. 123I crosstalk into the 99mTc window was 2.79% and was relatively constant with changes in the location of the 99mTc energy window. (b) Tomographic images: results showed 1.51% 99mTc crosstalk in the 123I image compared with 12.44% for the uncorrected image and 3.70% 123I crosstalk in the 99mTc image. CONCLUSION: An effective technique for the simultaneous acquisition of 99mTc and 123I radiotracer distributions in the brain has been developed and validated in a phantom model and should have clinical application in peri-ictal functional activation studies of the brain.

Comparative study of 99mTc-ECD and 99mTc-HMPAO for peri-ictal SPECT: qualitative and quantitative analysis.

OBJECTIVES: Most studies that clinically validated peri-ictal SPECT in intractable partial epilepsy had used technetium-99m-hexamethylpropylene amine oxime (99mTc-HMPAO or 99mTc-exametazime) as the radiopharmaceutical. Because of some theoretical advantages, technetium-99m-ethyl cysteinate diethylester (99mTc-ECD or 99mTc-bicisate) is increasingly being used instead. This study compares unstabilised 99Tc-HMPAO and 99mTc-ECD in the performance of peri-ictal SPECT in partial epilepsy. METHODS: The injection timing and localisation rates in 49 consecutive patients with partial epilepsy who had peri-ictal injections with unstabilised 99mTc-HMPAO were compared with 49 consecutive patients who had peri-ictal injections with 99mTc-ECD. Quantitative cortical/subcortical and cortical/extracerebral uptake ratios were also compared. Subtraction SPECT coregistered to MRI (SISCOM) was performed in patients whose interictal SPECTS were available. RESULTS: In the 99mTc-ECD patients, the latency from seizure commencement to injection was shorter (median 34 v 80 seconds, p<0.0001) and there was a lower rate of postictal injections (16.3% v 57.1%, p<0.0001). The cortical/extracerebral and cortical/subcortical uptake ratios were greater in the 99mTc-ECD images (median 5.0 v 3.6, and 2.5 v 2.2 respectively; both p<0.005), but the relative peri-ictal increase in uptake in the cortical focus did not differ significantly (median 37.0% v 37.0%; p>0.05). Blinded review of the SISCOM images were localising in a higher proportion of the 99mTc-ECD patients (40/45 (88.9%) v 25/37 (67.6%), p<0.05), and had a better concordance with EEG, MRI, and with the discharge diagnosis. CONCLUSION: 99mTc-ECD compares favourably with unstabilised 99mTc-HMPAO as a radiopharmaceutical for peri-ictal SPECT studies. Its use results in earlier injections and less frequent postictal injections than unstabilised 99mTc-HMPAO, thereby enhancing the sensitivity and the specificity of peri-ictal SPECT for the localisation of intractable partial epilepsy.

Subtraction SPECT co-registered to MRI improves postictal SPECT localization of seizure foci.

OBJECTIVE: To determine whether the detection of focal hypoperfusion by subtraction SPECT co-registered to MRI (SISCOM) improves the sensitivity and specificity of postictal SPECT in intractable partial epilepsy. BACKGROUND: Postictal SPECT injections are easier to perform than are ictal injections, but the images are more difficult to interpret and have been reported to have lower sensitivity and specificity. METHODS: Thirty-five consecutive intractable partial epilepsy patients who had postictal SPECT studies were evaluated. The following sets of SPECT images were separately interpreted by three blinded reviewers and classified as either localizing to 1 of 16 possible sites in the brain or as nonlocalizing: unsubtracted postictal and interictal images for conventional side-by-side comparison, SISCOM images of hyperperfusion, SISCOM images of hypoperfusion, and both sets of SISCOM hyperperfusion and hypoperfusion images (combined SISCOM evaluation). RESULTS: Significantly higher proportions of the hyperperfusion SISCOM images (65.7%), the hypoperfusion SISCOM images (74.3%), and the combined SISCOM evaluation (82.9%) were localizing than were the conventional method of side-by-side comparison of unsubtracted images (31.4%; p < 0.0001). Concordance with the discharge diagnosis was higher for the combined SISCOM evaluation than it was for either the hyperperfusion or the hypoperfusion SISCOM images alone (both p < 0.05). For the hypoperfusion SISCOM and the combined SISCOM evaluations, concordance of the localization with the site of epilepsy surgery was associated with a greater probability of an excellent outcome than were nonconcordant/nonlocalizing images (both p < 0.05). CONCLUSION: The use of SISCOM to detect focal cerebral hypoperfusion, in addition to focal hyperperfusion, improves the sensitivity and specificity of postictal SPECT in intractable partial epilepsy.

Optimized homomorphic unsharp masking for MR grayscale inhomogeneity correction.

Grayscale inhomogeneities in magnetic resonance (MR) images confound quantitative analysis of these images. Homomorphic unsharp masking and its variations have been commonly used as a post-processing method to remove inhomogeneities in MR images. However, little data is available in the literature assessing the relative effectiveness of these algorithms to remove inhomogeneities, or describing how these algorithms can affect image data. In this study, we address these questions quantitatively using simulated images with artificially constructed and empirically measured bias fields. Our results show that mean-based filtering is consistently more effective than median-based algorithms for removing inhomogeneities in MR images, and that artifacts are frequently introduced into images at the most commonly used window sizes. Our results demonstrate dramatic improvement in the effectiveness of the algorithms with significantly larger windows than are commonly used.

The practical utility of performing peri-ictal SPECT in the evaluation of children with partial epilepsy.

Peri-ictal brain single-photon emission computed tomography (SPECT) is increasingly being established as a useful test in localizing partial epilepsy in adults. However, obtaining an ictal injection and acquiring the SPECT images poses a greater challenge in pediatric patients, and few reports have specifically addressed the practical use of this technique in children. The Mayo Clinic experience of peri-ictal SPECT in the evaluation of children with partial epilepsy is reported here. Peri-ictal SPECT was attempted during 71 admissions involving 59 patients (median age 12 years, range 1 year 6 months-17 years). A peri-ictal SPECT injection was performed on 48 (67.6%) of these admissions in 43 (72.9%) patients, and only two patients could not be scanned. Of the 46 peri-ictal images successfully obtained, 30 (65.2%) were from ictal injection and 16 (34.8%) from post-ictal injections. Forty-two (91.3%) of the successfully obtained SPECT images, in 38 patients (92.3%), were classified as localizing (15 temporal, 24 extratemporal). We conclude that, with the appropriate unit setup and well-trained staff, peri-ictal SPECT scans can be obtained in most pediatric partial epilepsy patients. Moreover, the procedure provides specific localizing information in a high proportion of these patients.

Scalp-recorded EEG localization in MRI volume data.

Scalp-recorded EEG is a noninvasive and widely available tool for studying normal and dysfunctional human neurophysiology with unsurpassed temporal resolution. However, scalp-recorded EEG data is difficult to correlate with anatomy, and most current display and neural source estimation algorithms are based on unrealistic spherical or elliptical models of the head. It is possible to measure the positions of electrodes on the patient's scalp, and to register those electrode positions into the space of a high-resolution MRI volume, and to then use the patient-specific anatomy as the basis for display and estimation of neural sources. We use a surface matching algorithm to register digitized electrode and scalp surface coordinates to a three-dimensional MRI volume. This study uses fiducial markers in phantom and volunteer studies to quantitatively estimate the accuracy of the electrode registration method. Our electrode registration procedure is accurate to 2.21 mm for a realistic head phantom and accurate to 4.16 mm on average for five volunteers. This level of accuracy is considered within acceptable limits for clinical applications.

Subtraction ictal SPECT co-registered to MRI improves clinical usefulness of SPECT in localizing the surgical seizure focus.

Traditional side-by-side visual interpretation of ictal and interictal single-photon emission computed tomography (SPECT) scans can be difficult in identifying the surgical focus, particularly in patients with extratemporal or otherwise unlocalized intractable epilepsy. Computer-aided subtraction ictal SPECT co-registered to MRI (SISCOM) may improve the clinical usefulness of SPECT in localizing the surgical seizure focus. We studied 51 consecutive intractable partial epilepsy patients who had interictal and ictal scans. The SPECT studies were blindly reviewed and classified as either localizing to 1 of 16 sites in the brain or as nonlocalizing. SISCOM images were localizing in 45 of 51 (88.2%) compared with 20 of 51 (39.2%) for traditional side-by-side inspection of ictal and interictal SPECT images (p < 0.0001). Inter-rater agreement for two independent reviewers was better for SISCOM (84.3% versus 41.2%, kappa = 0.83 versus 0.26; p < 0.0001). Concordance of seizure localization with the more established tests was also higher for SISCOM. Late injection of the radiotracer (> 45 seconds), but not secondary generalization of the seizure, was associated with a falsely localizing or nonlocalizing SISCOM. Epilepsy surgery patients whose SISCOM localization was concordant with a falsely localizing or nonlocalizing SISCOM. Epilepsy surgery patients whose SISCOM localization was concordant with the surgical site were more likely to have excellent outcome than patients with nonconcordant or nonlocalizing findings (62.5% [10/16] versus 20% [2/10]; p < 0.05). On the other hand, seizure localization by the traditional method of SPECT inspection had no significant association with postsurgical outcome. We conclude that SISCOM improves the sensitivity and the specificity of SPECT in localizing the seizure focus for epilepsy surgery. Concordance between SISCOM localization and site of surgery is predictive of postsurgical improvement in seizure outcome.

Subtraction ictal SPET co-registered to MRI in partial epilepsy: description and technical validation of the method with phantom and patient studies.

Computer-aided subtraction of the co-registered and normalized interictal from the ictal single photon emission tomography (SPET) scan, followed by co-registration to the magnetic resonance image, may improve the utility of ictal SPET in the localization of partial epilepsy. This paper describes and technically validates our method. The SPET to SPET co-registration was tested using six sequential 99Tc(m) brain phantom SPET images of different known positions (15 matches). The registration error was determined by multiplying the calculated match transformation matrix by the inverse of the known transformation matrix. The 'worst case' co-registration error was less that one voxel diameter in all cases (median 3.2 mm, range 1.2-4.8 mm). For interictal to ictal SPET registrations in 10 consecutive intractable partial epilepsy patients, a similar root mean square distance (RMSD) between corresponding points on the matched scans was found as for the phantom studies (median 2.2 vs 2.6 mm). The appropriateness of our normalization was studied by comparing the pixel intensity distributions between the matched scans, and by analysing the subtraction pixel intensity distribution. The pixel intensity distribution for both the normalized phantom, and paired normalized patient studies, were closely matched to each other except for the extreme values, which in clinical situations likely represent regions of ictal activation or depression. The subtraction image intensity distributions were symmetrically centred on zero for all values up to at least within the 5th to 95th centile range, confirming good normalization for the 'non-activated' pixels. Also, a linear relationship was demonstrated between the measured pixel intensity on the phantom scans and the true changes in 99Tc(m) activity based on its decay constant. The results of this study demonstrate that our method produces accurate SPET to SPET co-registration, and appropriate SPET normalization, thereby allowing a valid ictal subtraction image to be derived.