A Technique for Automated Localization of Epileptic FociUsing SPECT Image Subtraction
Mark Rossman, M.S.C.E.,Malek Adjouadi, Ph.D., Melvin Ayala, Ph.D.
Ilker Yaylali, Ph.D., M.D. and Prasanna Jayakar, Ph.D., M.D.
Department of Electrical & Computer Engineering, FloridaInternationalUniversity
10555 W. Flagler Street, MiamiFL33174
Department of Neurology, Miami Children’s Hospital
3100 S.W. 62nd Avenue, Miami, Florida33155
Abstract: Accurate epileptic focus localization using SPECT (Single Photon Emission Computed Tomography) images has proven to be a challenging endeavor. Firstly, commonly used radio-pharmaceuticals such as hexamethylpropylene amine oxime (HMPAO) quantitatively underestimate large blood flows, leading to subtracted SPECT images that do not reflect the true cerebral physiological conditions, and often displaynon-distinct epileptic foci. The proposed relative change subtraction method of SPECT image analysishelps alleviate this quantitative burden. Secondly, the image analysis process traditionally performed by physicians is time-consuming and prone to error. Toward this end, an automated algorithm was designed to analyze SPECT images and provide feedback to users through a visual interface.
Key-Words: Epilepsy, HMPAO, Perfusion, SPECT Subtraction, Medical Images
1 Introduction
The main focus of this research is placed on the detection and accurate localization of epileptic foci using non-invasive imaging techniques. In the specific case of epilepsy surgery, the imaging capabilities provided through the integration of SPECT, Magnetic Resonance Imaging (MRI),and electro-encephalograms (EEG) are used to provide a reasonably accurate depiction of the in-vivo neurology, allowing vascular malformations, lesions, and abnormal perfusion and electrical patterns to be visualized. Since SPECT demonstrates the functional activity (bloodflow patterns) of the brain, it is consideredvaluable for diagnosing and localizing several neurological disorders, including epilepsy [1].
Because SPECT only provides static images of an organ, multiple SPECT studies must be created in order to accentuate its dynamics. Many researchers have agreed that epileptic foci are reliably localized with a subtraction SPECT image [2,3]. This image is created via the information from two individual studies, where one study depicts the brain when it is not presently undergoing an epileptic seizure (interictal SPECT) and the other depicts the brain during the epileptic event (ictal SPECT). This method is regarded as ‘basic subtraction’ and is still widely used in the literature as the current state of the art.
Areas of the brain that may be characterized as epileptogenic often display the following behaviors:
- During times between epileptic events, some brain tissue demonstrates that irritation exists through the discharge of electrical spikes. At the same time, this irritation also results in a reduced bloodflow in this tissue.
- Likewise, during seizure events, this tissue often experiences increased bloodflow.
In this study, areas of the brain that satisfied caveats (1) and (2) abovewere cited as “probable” epileptic foci by medical experts [4].
Subtraction between two SPECT scans is important because it accentuates the changes in cerebral perfusion of any pertinent regions apparent across the image pair, thus highlighting their locations. Though SPECT image subtraction is effective at demonstratingchanges in perfusion between SPECT imaging studies, there are some constraints to its utility. Firstly, the process of obtaining subtracted SPECT images requires many processing steps and is, in general, interactive, requiring inputs from users familiar with brain dynamics. To date, programming tools for the automation of SPECT image analysisarerarely mentioned in the specialized literature.It would be advantageous if a system were made available thatis fully automated. Secondly, the images obtained through SPECT subtraction do not always depict a localized epileptic focus. Toward alleviating these burdens, a SPECT analysis application was implemented in MATLABthat incorporates an improved SPECT subtraction method, relative change subtraction, and provides medical doctors with a turnkey solution with which to automatically derive useful diagnostic information using only the images themselves as inputs.
2 Materials and Methods
2.1 Patients
Ten patients with epilepsy were selected for this study by medical experts at Miami Children’s Hospital. Patients were selected that have had recurring epileptic seizures, intracranial EEG performed, and both ictal and interictal SPECT imaging studies performed. For the patients selected, ictal and interictal SPECT data were made available. Also, for each patient, a physicians’ report detailing the most probable epileptic focus location, based on observations made using intracranial EEG, was provided. These physicians’ assessments based upon intracranial EEGwere considered the “Gold Standard” by which the validity of the SPECT image analysis algorithm wasquantified. All of the procedures used in acquiring the necessary data followed strict protocols pursuant to the ethical guidelines and regulatory requirements regarding research involving human subjects.
2.2 Radiopharmaceutical, SPECT Imaging System and Acquisition Parameters
For neurological studies, the pharmaceutical HMPAO (hexamethylpropylene amine oxime) bound to Tc99mwasused because it has been shown to be effective in demonstrating cerebral blood perfusions [5,6]. Likewise, the efficacy of HMPAO as an imaging tracer for epileptic studies is well accepted by the medical imaging community. The administered dose of Tc99m-HMPAOwas compensated to a level between 5mCi (185 MBq) and 20mCi (740 MBq) using the patient weight (500 Ci/kg). After the dosage was intravenously administered, each patient was then required to wait for a period of 90 minutes prior to the image acquisition procedure to allow for proper uptake of the pharmaceutical.
A Siemens MULTISPECT3 triple-head gamma camera imaging system (Siemens, USA) wasused to acquire the SPECT images. Each patient, after injection and waiting period, was scanned in a dimly-lit room with eyes closed,using a predefined protocol.
3 Automated System for SPECT Image Analysis
The contribution of this study was an automated SPECT image analysis algorithm, which included all of the steps shown in Figure 1.
Figure 1: Automated SPECT processing algorithm
In the following subsections, a detailed description of the algorithm steps will be made.
3.1 Image Registration
Registration is a necessary image pre-processing task whenever it is desired to compare two or more images that have been created from different physical viewpoints or using different imaging systems. Thus, for the purposes of this study, before meaningful subtractive SPECT images could be obtained, ictal and interictal images were co-registered to reduce the effects of patient motions between imaging sessions.
It is generally accepted that affinetransformations can successfully correct for any study misalignments between intra-patient, intra-modality images [7,8]. Toward this end, an intensity-based registration algorithm was implemented that employs the affine transformation model with the optical flow error as the similarity measure. The mathematical strategy to determine the optimal coefficients that correctly registered the image pair consisted of minimizing the optical flow error between the image pair using spatio-temporal derivatives [9,10]. It was shown through a previous validation study that the registration results produced by the proposed algorithm were accurate enough for intra-patient SPECT brain studies and gave comparable registration accuracies to those provided by professionally-produced software packages [11].
3.2 Image Intensity Normalization
Knowledge of the radioactive distribution differences between imaging studies is important [12]. Since subtractive SPECT involves image pixel intensity differences, the values of these differences arepartially dependent on the amount of radioactivity in each study. To eliminate this activity difference, the studies were first globally normalized [2]. For normal SPECT scans (acquired from patients without brain disorders), the ratio of mean intensities may be used as a metric for the activity differences between two images. However, for the case of SPECT scans of epileptic brains, other image intensity normalization metrics are normally recommended. A method that has produced reliable resultscalculates the optimal normalization factor using the stochastic sign change (SSC) criterion[13]. Since it produced better accuracy, the SSC method of image normalization was used in this study.
3.3 Artifact Reduction
Artifact reduction was performed by using a simple thresholding technique. A thresholding example is shown in Figure 2, where the reader can observe the influence of the applied percentage on the quality of the artifact reduction.
Figure 2: SPECT image after thresholding
3.4 Image Subtraction Methods
To illustrate changes in cerebral perfusion between ictal and interictal brain states, the SPECT images were subtracted. Basic subtraction consists of applying the operation ‘ictal minus interictal’ to the two individual SPECT imaging studies. Even though basic subtraction is often effective, it sometimes lacks specificity toward isolation of epileptic foci, even after proper image normalization has taken place. To lessen this unwanted effect, an experimental subtraction method was proposed through research efforts involving the Miami Children’s Hospital Neuroscience Group. Percent change in perfusion, or relative change subtraction, is the mathematical operation ‘basic subtraction divided by interictal’as in equation (1).This augmented subtraction method is an attempt to normalize the subtractive SPECT intensities so that the percentage change in blood perfusion between the studies can be observed. In previous studies, this method of subtraction demonstrated itself to have a better sensitivity and specificity for epileptic focus localization than basic subtraction [11]. This augmented the ability of the subtracted image to accurately pinpoint the epileptic focus, increasing the utility of the image as a diagnostic tool.
(1)
Contrasting results of basic versus relative change subtraction are as shown in Figures 3 and 4.
Figure 3: Program outcome after applying the basic SPECT subtraction method to a patient
For the patient case shown in Figure 3, a very broad focus exists. It is C-shaped and extends from the left temporal/parietal to the left frontal/temporal region. It is apparent that this case does not localize effectively to a small number of image slices with basic subtraction. The focus mentioned in Figure 3 also appears in the relative change subtraction image in Figure 4. However, in Figure 4, the most intense portions (dark gray, black) only seem to concentrate in the central left temporal/parietal region in slices #34 - #37. Thus, in this case, relative change subtraction improved the localization accuracy of the SPECT images to a fewer number of distinct slices: 4 using relative change subtraction (slices 34-37) versus 8 to 9 using basic subtraction(slices 29-37).
Figure 4: Program outcome after applying the relative SPECT subtraction method to a patient
Due to the effectiveness of relative change subtraction and its ability to produce better results than basic subtraction, only this method of subtraction was used in the automated SPECT analysis algorithmpresented in this study.
3.5 Image Intensity Scaling and Color Display
Grayscale display techniques are usually employed in the traditional SPECT image interpretation process. Since the human eye can only differentiate a relatively small number of gray levels, the contrast of these images may appear limited[14]. A proposition was to display images with pixel intensities made proportional to a color map, as shown in Figure 5. This type of display is not only more appealing but also helps in the quantitative image interpretation process.The main window of the application is shown in Figure 6. Interaction with the main window will activate additional windows that show the subtracted SPECT images in full color.
Figure 5: SPECT multi-color display application.
Figure 6: Main window of the application. This window is used for file opening and options selection. It also displays the names of the proposed regions along with additional windows that show the processed images.
3.6 Epileptic Focus Localization Method
Once a subtracted SPECT image was created using the aforementioned steps, determining the location of the epileptic focus remained. An epileptic focus is generally defined as a concentrated group of image pixels whose values have a similar high intensity, which is analogous to high changes in perfusion. Generally, thresholding can be used to separate foci pixels from background pixels.
Once a focus is isolated through thresholding, it needs to be localized and assigned a location in the brain. This technique used nomenclature that was consistent with the major regions of the brain (frontal, temporal, parietal, or occipital lobes). A simplified two-dimensional model of the brain was thus proposed, as illustrated in Figure 7:
Figure 7: Simplified 2-D brain model
The above representation of the brain was created to subdivide it into 6 major anatomical regions as follows: Region 1: Right Frontal Lobe, Region 2: Left Frontal Lobe, Region 3: Right Temporal Lobe, Region 4: Left Temporal Lobe, Region 5: Right Occipital/Parietal Lobe, Region 6: Left Occipital/Parietal Lobe.
Focus localization in three-dimensional subtracted SPECT imagesinvolves the following tasks:
- Using the pixels of the mid-axial 2-D slice image, calculate the values Xmid, Ylower, Yupper.
- Threshold the 3-D subtracted SPECT image.
- Find the coordinates (x,y,z) of all remaining nonzero image voxels.
- Using the (x,y) coordinate of each voxel and the values of Xmid, Ylower, and Yupper, determine to which region (1-6) it belongs.
- Count the total number of voxels in each region.
This set of tasks produced results similar to those shown in Figure8 and Table I. For each threshold value, a pixel count of the image by region was performed. Based on these pixel counts, a determination about which region(s) had the highest pixel populations was made.
Figure 8: Two SPECT Images afterapplication ofsteps 1 trough 4 of the localization procedure
Table I: Pixel count of SPECT images in Figure 8
Region / Case 1 / Case 2# pixels in region 1 / 48 / 9
# pixels in region 2 / 6 / 11
# pixels in region 3 / 14 / 63
# pixels in region 4 / 8 / 58
# pixels in region 5 / 5 / 7
# pixels in region 6 / 5 / 13
Total # of pixels / 86 / 161
Overall focus region / Region 1 / Regions 3, 4
As the algorithm completed its tasks, it produces a description of the location of the seizure activity. Since the localization result is dependent on the value of image intensity threshold used, this value is also included in the results. Shown in Table II is an example of this focus localization procedure using one set of patient SPECT data. For this case, the automated algorithm predicted region 4 (left temporal lobe) as the brain region with the highest probability of containing the epileptic focus.
In Table II, threshold range 1 is 50% for basic subtraction and 5% for relative change subtraction, respectively. For ranges 2, 3 and 4 the values are 60% and 10%, 70% and 15%, and 80% and 20%, respectively. The last percentage values (5%, 10%, 15% and 20%) were found to yield the most significant changes in the overall results[11]; therefore, these values werethe ones used for relative change subtraction in this study.
Table II: Seizure focus localizations for the case shown in Figures 3 and 4(: threshold mean, : threshold standard deviation).
Patient No. 2 / Basic Subtraction / Relative Change SubtractionRange 1 / 4 / 4
Range 2 / 4 / 4
Range 3 / 4 / 4
Range 4 / 4, 6 / 4
+ 1 / 4, 6 / 4
+ 1.5 / 4 / 4
+ 2 / 4 / 4
+ 2.5 / 4 / 4
Vote per region / 8 for reg. 4
2 for reg. 6 / 8 for reg. 4
Vote count / 10 / 8
Vote Percentage / 80% for reg. 4
20% for reg. 6 / 100% for reg. 4
4 Algorithm Performance Analysis
4.1 Receiver Operating Characteristic
The localization results provided by the algorithm were gauged against expert medical assessments for each case using receiver operating characteristic (ROC) analysis. This analysis gave a quantitative result as to the sensitivity, specificity, and accuracy of the proposed algorithm.
For each patient, the six regions of the brain existed as separate entities to be tested for the presence or absence of an epileptic focus. Since there wereten patients, this gave60 separate entities to be tested. The algorithm gave the following answers: 0 (indeterminate region), 1-6, or any combination of two adjacent brain regions (1,2; 1,3; 2,4; etc.). Likewise, the doctors’ assessment consisted of a single brain region (region 1, 2, 3, 4, 5 or 6) that contained the epileptic focus. As stated earlier, the physicians’ assessmentswere derived from the visual inspection of intracranial EEG obtained from each patient. The ten individual brain regions (1 focus region for each patient) were the only “positive” or “abnormal” brain regions; the remaining 50 individual brain regions wereautomatically declared as “negative”regions. Based on this testing scenario, the sensitivity, specificity, and accuracy of the automated SPECT analysis algorithm were calculated for different threshold values (Table III). Such positive outcomes could be integrated with EEG- or MRI-based analysis for added validation [15, 16].
Table III: Sensitivity, specificity and accuracy of automated SPECT analysis as a function of threshold value (: threshold mean, : threshold standard deviation)
Thresh. / TPF(sensitivity) / TNF
(specificity) / Accuracy
5% / 0.67 / 0.93 / 0.89
10% / 0.83 / 0.97 / 0.94
15% / 0.83 / 0.97 / 0.94
20% / 0.83 / 0.97 / 0.94
+ 1 / 0.50 / 0.90 / 0.83
+ 1.5 / 0.83 / 0.97 / 0.94
+ 2 / 0.67 / 0.97 / 0.91
+ 2.5 / 0.67 / 0.97 / 0.91
Avg. values / 0.728 / 0.956 / 0.913
It can be observedfrom Table III that the highest parameters corresponded to the thresholds 10%, 15%, 20%, as well as the Mean + 1.5 Std threshold.
Table IV: Physician assessments and final results obtained after running the system with available data
Patient # / Phys. Focus / 5% / 10% / 15% / 20% / +1 / +1.5 / + 2 / + 2.5 / Agreement1 / 3 / 3 / 3 / 3 / 3 / 3 / 3 / 3 / 3 / 100%(3)
2 / 3 / 3 / 3 / 3 / 3 / 3 / 3 / 3 / 0 / 87.5% (3)
12.5% (0)
3 / 3 / 3 / 3 / 3 / 1 / 3 / 3 / 1 / 1 / 62.5% (3)
37.5% (1)
4 / 3 / 3 / 3 / 3 / 0 / 3 / 3 / 3 / 3 / 87.5% (3)
12.5% (0)
5 / 2 / 2 / 2 / 2 / 2 / 2 / 2 / 2 / 2 / 100% (2)
6 / 4 / 4 / 4 / 4 / 2 / 4 / 4 / 4 / 2 / 75% (4)
25% (2)
7 / 4 / 4 / 4 / 2 / 2 / 4 / 4 / 2 / 2 / 50% (4)
50% (2)
8 / 1 / 1 / 1 / 1 / 0 / 1 / 1 / 1 / 1 / 87.5% (1)
12.5% (0)
9 / 2 / 2 / 2 / 2 / 0 / 2 / 2 / 2 / 0 / 75% (2)
25% (0)
10 / 1 / 1 / 1 / 3 / 3 / 1 / 1 / 1 / 3 / 62.5% (1)
37.5% (3)
The aforementioned range of thresholds was used in the form of a committee (as shown in Table IV) to eliminate the dependence of the system performance parameters on a single threshold value. Specifically, this study uses all the thresholds shown in Table III in a committee and attempts to find a solution by using the contribution of each one of the thresholds to create a composite result, so as to consolidate the region-based agreement. Results of this process are shown in Table IV.
In Table IV, the 2nd column represents the regions declared by the physiciansto contain the epileptic focus, whereas the following eight columns show the predominant region,which wasdetermined using each one of the individual thresholds. The algorithm then summarizes the results (last column) by counting the number of occurrences of each detected regionand dividing it by the total number of possible detections (8). This percentage of agreement between the threshold-based committee members is the final output of the proposed algorithm.