WO2007061807A2 - Processing of functional imaging data - Google Patents

Abstract

Techniques for implementing temporal dynamic data processing are disclosed. In one aspect, functional imaging data is processed by receiving the functional imaging data and separating the received data into multiple components. One or more of the separated components are selected based on a threshold, and the selected components are spatially normalized. Based on one of the selected and spatially normalized components, grouped regions of activity (ROAs) are constructed. In the constructed grouped ROAs, one or more Brodmann areas (BAs) are determined. Based on the determined BAs, a temporal dynamic pattern of activation is determined.

Description

PROCESSING OF FUNCTIONAL IMAGING DATA

[0001] This application claims the benefit of the filing date of U.S. Patent Application No. 60/737,983, filed on November 17, 2005, the contents of which are incorporated by- reference as part of this application.

TECHNICAL FIELD

[0002] This application relates to processing of functional imaging data such as Functional Magnetic Resonance Imaging (fMRI) data.

BACKGROUND

[0003] Functional imaging techniques such as Functional Magnetic Resonance Imaging (fMRI) has typically been used to observe various brain responses and activities and to diagnose certain brain conditions. In general, analysis of fMRI data has tended to implicate certain areas of the brain (e.g., somatosensory cortex, cingulated gyrus, limbic system) in central processing of pain while leaving the role of other areas of the brain (e.g., prefrontal cortex, insular and cerebellum) largely unsettled. For the most part, analysis of fMRI data has typically been based on experimental time-course driven data analysis methods such as Statistical Parametric Mapping (SPM) or Analysis of Functional Neurolmages (AFNI) . Such time-course driven data analysis methods tend to provide very little information about the dynamics of central signal processing. In addition, implementations of SPM and AFNI typically adopt multiple levels of linear correlation in imaging data analysis and tend to under assess the regions of activities (ROAs) in the brain that may be crucial in understanding the overall central processing of different stimuli . SUMMARY

[0004] Techniques for processing functional imaging data using temporal dynamic processing is disclosed. Implementations of a method, product and system as described in this specification may include various combinations of the following features.

[0005] In one aspect, functional imaging data is processed by receiving the functional imaging data and separating the received functional imaging data into multiple components. One or more of the separated components are selected based on a threshold, and the selected components are spatially normalized. Based on at least one of the selected and spatially normalized components, grouped regions of activity (ROAs) are constructed. In the constructed grouped ROAs, one or more Brodmann areas (BAs) are determined. In addition, temporal activities of the determined BAs can be constructed. [0006] Implementations can optionally include one or more of the following features. The functional imaging data can be received from a functional magnetic resonance imaging (fMRI) system. Alternatively, the functional imaging data can be received from a non-MRI system. For example, a non-MRI system can include a functional positron emission tomography (PET) system. A non-MRI system can also include a functional single photon emission computer tomography (SPECT) system. A non-MRI system can further include a functional near infrared imaging (NIR) system. The received functional imaging data can be separated into multiple components, with each component having a unique time course and corresponding regions of activity (ROAs) using independent component analysis (ICA) . Other suitable linear or nonlinear decomposition algorithms, such as (but not limited to) principal component analysis (PCA) , factor analysis (FA), etc., can also be used to separate recorded functional neuroimaging data into components. [0007] Implementations can also optionally include one or more of the following features. The functional imaging data can be received from each subject by measuring one or more baseline sensory thresholds of each human subject and capturing one or more functional images of the human subject while applying stimuli to the human subject. The stimuli applied while capturing the functional images can be based on the measured baseline sensory thresholds. The functional images can be captured while determining a time course of the human subject indicating an onset of pain.

[0008] Further, implementations can optionally include one or more of the following features. Post-ICA components that are related to the stimulus (e.g., hot pain) of cognitive task can be selected by correlating the time course of each component with the time course of the human subject indicating an onset and offset of the stimulus such as thermal pain which is delivered at a predetermined subject specific pain threshold. The correlation can be used to determine a linear correlation coefficient for each component. Other discriminate functions, such as (but not limited to) linear distance measure, group distance, linear or nonlinear regression, etc., can also be used to establish the relationship between decomposed components and the human subject responses and thus to select the most relevant component (s) . The component selection can be based on a preset threshold of linear correlation coefficiency. Also, individual ROAs specifically related to the stimulus can be constructed by spatially normalizing the component which consists of the highest level of linear correlation coefficient with the behavioral time course. In addition of a group ROA can be constructed by combining the individual components with the highest linear correlation coefficiency from each subject. Further, in each BA, a total number of voxel clusters present can be determined, and based on the determined total number of voxel clusters, an output can be generated. The generated output can include one or more three-dimensional surface contour plots of the relevant BAs.

[0009] The subject matter described in this specification can be implemented as a method or as a system or using computer program products, tangibly embodied in information carriers, such as a CD-ROM, a DVD-ROM, a semiconductor memory, and a hard disk. Such computer program products may cause a data processing apparatus to conduct one or more operations described in this specification.

[0010] In addition, the subject matter described in this specification can also be implemented as a system including a processor and a memory coupled to the processor. The memory may encode one or more programs that cause the processor to perform one or more of the method acts described in this specification.

[0011] In various aspects, ROAs involved in thermal pain tend to include bilateral primary and secondary somatosensory cortex, motor cortex, anterior cingulate gyrus and thalamus. These ROAs have been mostly implicated in pain processing typically in one or more combinations only. Identification of all of these ROAs in a single study has mostly been unsuccessful . Further due to factors such as limitations in pain study design, deficiencies of conventional analytical methods and lack of behavioral correlation, the regions of activities (ROAs) in the brain specific for decoding and modulating noxious signals typically have not been well defined. More over, the temporal dynamic relationships of these ROAs, which are important in understanding how the brain processes pain in a particular disease state, are largely unknown.

[0012] In some implementations, functional imaging dynamic signal processing (FIDSP) avoids some of the pitfalls of the conventional functional neuroimaging data processing. FIDSP provides for dynamically processing functional neuroimaging data such as functional MRI, functional PET, functional SPECT (single photon emission computer tomography) , functional NIR (near infrared) image, etc.. FIDSP can also be implemented to obtain information too difficult to extract or not available in conventional functional imaging data processing techniques. [0013] Particular implementations of the techniques described can be implemented to realize one or more of the following advantages. For example, implementations of FIDSP can be used to achieve a higher specificity in identifying the regions of activities (ROAs) related to applied stimuli than data processed by general linear model . In addition, FIDSP can be highly sensitive in identifying potential ROAs that may be associated with the stimuli or tasks. In particular, FIDSP can be implemented to use a data-driven analytical method and time course obtained from subject behavioral response to overcome the shortcoming of conventional experimental time- course analytical methods. Temporal dynamic information can be obtained from functional imaging measurements and presented in an easily understandable manner.

[0014] The details of one or more implementations of the techniques are set forth in the accompanying drawings, the description, and the claims. Other features, aspects, and advantages of the techniques will become apparent from the drawings, the description and the claims.

DESCRIPTION OF DRAWINGS

[0015] FIG. 1 is a functional block diagram of a FIDSP analysis system.

[0016] FIGS. 2A and 2B are flowcharts for performing FIDSP analysis on functional imaging data.

[0017] FIG. 3 is an image of the brain depicting stimuli- related activation of relevant grouped ROAs.

[0018] FIG. 4 is a template of Brodmann' s Area (BA) .

[0019] FIG. 5 shows 3-D surface contour plots for BAs 1-9- [0020] FIG. 6 shows 3-D surface contour plots for the Prefrontal Association Cortex (BAs 9-12) .

[0021] FIG. 7 shows 3-D surface contour plots for the Limbic Association cortex (BAs 23-33) .

[0022] FIG. 8 shows 3-D surface contour plots of the prefrontal cortex (BAs 45-46) .

[0023] FIG. 9 illustrates 3-D surface contour plots of a bilateral cerebellum.

[0024] FIG. 10 shows 3-D surface contour plots illustrating and interpreting the visual cortex (BAs 9-12) .

[0025] FIGS. HA, HB, HC, HD, HE and HF are flowcharts for implementing FIDSP data analysis.

[0026] FIG. 12 is a sample log folder.

[0027] FIG. 13 is a sample log file.

[0028] FIG. 14 is a sample log.txt file.

[0029] FIG. 15 is a sample terminal window.

[0030] FIG. 16 is a sample argument command.

[0031] FIG. 17 is a sample MATLAB^® screen.

[0032] FIG. 18 is a sample end screen.

[0033] FIG. 19 is a sample final MATLAB^® screen.

[0034] Like reference symbols in the various drawings indicate like elements.

DETAILED DESCRIPTION

[0035] Techniques for dynamically processing functional imaging data (e.g., fMRI measurements) using Functional Imaging Dynamic Signal Processing (FIDSP) are disclosed. In particular, FIDSP is implemented to provide temporal dynamic processing of thermal pain.

[0036] Various fMRI processing methods tend to rely mainly on an experiment time-course driven Gaussian model with multiple levels of linear correlation analysis. Blindly using experimental time-course for data analysis could potentially underestimate the ROA that may be relevant in pain processing and modulation. FIDSP can be used to implement a data driven approach that takes advantage of correlation to each subject's behavioral feedback time-course (e.g., time course of subject's button pressing in response to stimuli) . Under FIDSP, variation in latency of each subject's response (button-press) to noxious thermal stimulation is noted. [0037] In one aspect, PIDSP applies Independent Component Analysis (ICA) to functional imaging (e.g., fMRI) in such a way that improves the specificity, resolution and information that can be derived from analysis of stimuli (e.g., pain) that are centrally-processed by the brain. Implementations of FIDSP can be used to identify task/stimulus specific ROAs in the brain with a very high sensitivity and specificity without presuming when and where the activities may occur. The ROAs can be identified to provide pertinent information regarding regions of specific temporal dynamic sequence of activities. This information can be used to generate a unique brain processing fingerprint (or mapping) in response to a specific task or stimulus, and to allow easy interpretation of fMRI data.

[0038] FiG. 1 is a functional diagram of a system 100 for dynamically processing functional imaging data (e.g., fMRI data) using FIDSP. The system 100 can include a computer system 110, a display device 120, an input device 130, a storage device 140, a Thermal Analyzer 150, and a functional imaging (e.g., fMRI) system 160. The computer system 110 can include at least a processor 112 and a memory 114. The processor 112 can include a central processing unit (CPU) or other suitable processor/hardware such as an application specific integrated circuit (ASIC) . The memory 114 can be a volatile and/or non-volatile memory unit used to store and execute one or more processes for dynamically processing fMRI data.

[0039] The computer system 110 can be communicatively coupled to the display device 120, the input device 130, and the storage device 140 through appropriate communication channels 122, 132, and 142. In addition, the computer system 110 can be communicatively coupled to the thermal analyzer 150 and/or the functional imaging system 160 through appropriate communication channels 152 and 162 respectively. The communication channels 122, 132, 142, 152, and 162 can provide either unidirectional or bidirectional communications, and can be combined into a single, shared communication channel (e.g., a bus network) . Further, the communication channels can be implemented using either wired or wireless medium. Examples of wired communication channels include various Universal Serial Bus (USB) and FireWire connections. Examples of wireless communication channels include Bluetooth, WiFi, and WiMAX. In some implementations, the system 100 can be implemented using a combination of computer hardware and software .

[0040] The computer system 110 can be implemented to control and operate the thermal analyzer 150 in obtaining thermal sensory threshold measurements for each of the selected human subjects. Based on the measured thresholds for each subject, the computer system 110 can be implemented in conjunction with the functional imaging system 160 to obtain fMRI image data, for example. The received fMRI data is processed using the computer system 110 to perform temporal dynamic processing as described with respect to FIG. 2 below. [0041] FIG. 2 is a flow chart 200 depicting a process for dynamically processing fMRI data using FIDSP. With Institutional Review Board (IRB) approval, healthy human subjects are recruited at 210 for a study based on the inclusion and exclusion criteria listed in Table 1. Table 1. Criteria for Inclusion S- Exclusion of Human

Subjects

Pre-scan thermal threshold measurement

[0042] For each human subject, baseline sensory thresholds (e.g., warm, cold, cold pain and hot pain) are measured at 220. To preserve consistency in the location of threshold measurements among the subjects, a measurement template is used to mark the measurement location on each subject. The measurement template, for example, can be an elastic band having incremental markings. In some implementations, a location between the 6^th and 7^th markings of an elastic band having a total of 13 increments, extending from the medial malleolus to the medial tibial plateau is selected for the site of measurement.

[0043] The baseline sensory thresholds for each subject is measured using a Peltier Thermal Analyzer (Medoc Advanced Medical Systems, Durham NC) or other suitable thermal analyzers. A Peltier Thermal Sensory Analyzer includes a thermode stimulator of various dimensions (e.g., measuring 46 X 29 mm) designed to deliver various thermal stimuli. The temperature of the thermode stimulator (and the stimuli generated by the thermode stimulator) can be incrementally increased or decreased (e.g., at a rate of 1.2 degrees Celsius/sec for cold and warm sensations, and 3 degrees Celsius/sec for cold and warm pain) from a baseline of 32 degrees Celsius, depending on the sensation tested. [0044] During threshold measurements, each human subject is provided with a switch, a button or other suitable input devices to allow the subject to indicate, by interacting with the input device (e.g., by pressing a button), the onset and/or offset of the tested sensation (e.g., feeling of cold or warm pain) . Pressing the button not only signals the onset and/or offset of pain but also reverses the temperature change of the applied stimuli until a baseline temperature of 32 degree Celsius is reached. The computer system 110 can be implemented to record the temperature of the thermode (and the applied stimuli) at each time the button is pressed by the subject. In addition, the recorded temperatures for the tested stimuli are averaged by the computer system 110, and the generated average values displaced on the screen. The desired stimuli is repeated for a predetermined number of times and averaged. For example, a cold pain stimulus can be applied three times and the temperature at each time the button is pressed by the subject averaged. The average temperature values can be generated automatically by the computer system 110 or manually in response to a user input received through the input device 130.

[0045] Such peripheral sensory testing is further described in various studies. (See, e.g., Leung, A, Khadivi , B, Duann, JR, Cho, ZH and Yaksh, T, The effect of Ting point (tendinomuscular meridians) electroacupuncture on thermal pain: a model for studying the neuronal mechanism of acupuncture analgesia, J Altern Complement Med, 11 (2005a) 653-61; Leung, A, Wallace, MS, Ridgeway, B and Yaksh, T, Concentration-effect relationship of intravenous alfentanil and ketamine on peripheral neurosensory thresholds, allodynia and hyperalgesia of neuropathic pain, Pain, 91 (2001) 177-87; and Leung, AY, Wallace, MS, Schulteis, G and Yaksh, TL, Qualitative and quantitative characterization of the thermal grill, Pain, 116 (2005b) 26-32.)

FMRI Scanning

[0046] The determined sensory thresholds, represented as average temperature values, are used to apply subject-specific stimuli to each subject while capturing various fMRI images at 230. By applying subject specific noxious thresholds as stimuli, the possibility of either over-stimulating or under- stimulating the subjects is minimized. Further, application of subject specific stimuli provides increased sensitivity and specificity in identifying the ROA related to pain processing and modulation.

[0047] Each subject is placed comfortably in a supine position in the scanner with their eyes protected by an eye shield. Axial and lateral head restraints are also applied to minimize head movement. Hot pain stimuli with subject- specific thresholds are delivered to each subject's (N=IO subjects) medial calves for 15 seconds followed by a baseline temperature (32⁰C) stimulus for 60 seconds. The sequence of pain stimulation is repeated multiple times (e.g., four times) . Again, each subject is provided with a switch, a button or other suitable input devices to allow the subject to indicate (e.g., by pressing the button) the onset and/or offset of applied pain stimuli. The times at which the button is pressed by each subject are used to construct a time course of button pressing (onset and/or offset of pain) for each subject. The fMRI images of each subject can be obtained using various conventional MRI scanners. For example, a 3T GE scanner (GE Healthcare, United Kingdom) with 12*- weighted EPI-sequence (TE=30 ms, TR=2.5s, α=90°, TH=4mm, 30 slices, FOV=220x220 mm², MA=64x64) can be implemented. Two Tl-weighted images are acquired: one for spatially normalizing the functional images and the other one for anatomical details. [0048] Further, the thermal stimuli applied to each subject is provided in an oscillating pattern to minimize the possibility of "wind-up" of pain.

Functional Dynamic Signal Processing (FIDSP)

[0049] The fMRI image data acquired at 230 is received from a functional imaging system 160 such as a fMRI system and FIDSP is performed on the acquired data at 240. In general, fMRI image data is made up of thousands of volume elements (e.g., voxels), which represent not only the desired task (stimuli) -related changes (e.g., activations) in the brain but also includes nontask-related activations and motion and/or machine artifacts. In order to isolate the task-related activations, the acquired fMRI image data is processed to separate or unmix into individual components using Independent Component Analysis (ICA) at 242. ICA is a computational process for separating a multivariate signal into additive subcomponents assuming statistical independence of non- Gaussian source signals. The process of implementing ICA is further described in various studies. (See, e.g., Duann et al . 2002; McKeown et al . 1998.) Alternatively, other suitable linear or nonlinear decomposition algorithms, such as (but not limited to) principal component analysis (PCA) , factor analysis (FA), etc., can also be used to separate recorded functional neuroimaging data into components. [0050] The captured fMRI data can be unmixed or separated into independent components using fMRLAB (University of California at San Diego, San Diego CA) or other suitable functional imaging data analysis tools. FMRLAB is a Matlab toolbox for fMRI data analysis that applies ICA. FMRLAB is typically implemented using LINUX, but FMRLAB is compatible with other suitable platforms (e.g., Unix, Widows , etc.) By using fMRLAB, fMRI image data can be decomposed or separated into individual independent components, each component having a unique time course and corresponding regions of activities (ROAs) .

[0051] Since not all of the separated components are based on task (e.g. stimuli) -related activations in the brain, only the components of interest (those related to task-related activations) are selected based on a preset threshold linear correlation coefficiency at 244. Post-ICA components that are related to the stimulus (e.g., hot pain) of cognitive task can be selected by determining a linear correlation coefficient for each component by correlating the time course of each component with the time course of the human subject indicating an onset and/or offset of the stimulus such as thermal pain which is delivered at a predetermined subject specific pain threshold. Other discriminate functions, such as (but not limited to) linear distance measure, group distance, linear or nonlinear regression, etc., can also be used to establish the relationship between decomposed components and the human subject responses and thus to select the most relevant component (s) . For example, one or more components of interest can be selected from, the separated components based on a preset threshold of linear correlation coefficiency between the time-course of each component and each subject's button- press time-course.

[0052] The selected components of interest are used to construct a temporal dynamic pattern of activation. For each subject, the components selected had a linear correlation coefficient greater than a preset threshold of 0.3. In some implementations, the preset threshold of the linear correlation coefficiency is varied (e.g., increased) based on the desired data processing. Increasing the threshold too much may eliminate some of the relevant data, and decreasing the threshold too much may include some of the non-relevant data.

[0053] The selected components for each subject are normalized at 246. In particular, spatial normalization of the selected components are performed based on the Montreal Neurological Institute standard. In some implementations, other suitable applications for spatially normalizing functional imaging data are implemented. For example, Free Server spatially normalizes to provide a mapping of the cortical region. In neuroimaging, spatial normalization is an image processing method, more specifically an image registration method. Human brains differ in size and shape, and spatial normalization allows the human brain scans (the functional imaging data) to be deformed so that one location in one subject's brain scan corresponds to the same location in another subject's brain scan. Such deformation in spatial normalization enables identification of common brain activation across multiple human subjects. The functional imaging data can be obtained from magnetic resonance imaging (MRI) or positron emission tomography (PET) scanners. [0054] Spatial normalization can include two processes: (1) specification/estimation of warp-field and (2) application of warp-field with resampling. The estimation of the warp-field can be performed in one modality, e.g., MRI, and be applied in another modality, e.g., PET, if MRI and PET scans exist for the same subject and they are coregistered. Spatial normalization typically employ a 3 -dimensional non-rigid transformation model (a "warp-field") may be parameterized by basis functions such as cosine and polynomia. There are a number of programs that implements both estimation and application of a warp-field, which can be found as a part of statistical parametric mapping (SPM) programs.

[0055] From the selected components, for each subject, the component having a time-course with the highest level of linear correlation with the corresponding button press time- course is used for constructing an initial grouped ROA at 248. The significant voxel cluster, which consists of the initially grouped ROA and a spatial extent of 10 voxels in all directions with a z value > 3.5, normally corresponds to an uncorrected P<0.001. However, since the distribution of the components as shown on component maps are super-Gaussian, the uncorrected p value is significantly less than 0.001.

[0056] The initially grouped ROA is then processed to identify corresponding Brodmann Areas (BAs) at 250, which can be used as areas of interest for constructing temporal dynamic relationships. A Brodmann area is a region in the brain cortex defined in many different species based on each specie's cytoarchitechure, the organization of the cortex as observed when a tissue is stained for nerve cells. The BAs are identified by transforming the individual subject's fMRI data and resultant component maps into a standard brain template (e.g., as provided by Montreal Neurology Institute), which includes the definition of BAs associated with it.

[0057] For each subject, time shifts one for each selected component needed to generate a maximum linear correlation between the time-course of each selected component and the subject's button press time-course is determined at 252. The determined time shift for each selected component is used as a reference for the relative time course in determining the temporal dynamic or sequence of events.

[0058] For each selected component at a relative time point

(e.g., the reference for the relative time course and thus the time shift for each selected component) , the number of voxel clusters (positively and negatively correlated) in each relevant BA is recorded for each subject at 254. The relevant BAs of ROA are the components having the highest linear correlation in terms of their time courses.

Temporal Dynamic Processing of Pain

[0059] Based on the different degrees of component time shift obtained from the maximal correlation for each individual analysis, a temporal dynamic pattern of activation or deactivation in each BA is determined at 256. The temporal dynamic analysis of the ROAs related to pain shows the presence of similarities and differences in the activation pattern among different ROAs in response to noxious thermal pain stimulations. For example, the temporal dynamic analysis shows a similar temporal dynamic activation pattern in bilateral primary, secondary and tertiary somatosensory cortices. However, the temporal dynamic activation pattern between the left and right premotor and motor cortex, which can be due to the anticipatory motor response to noxious stimuli, are shown to be different.

[0060] In addition, the temporal dynamic activation analysis shows a slight delay in the activation of prefrontal and limbic association cortex in relationship to the somatosensory cortex response to noxious thermal stimuli. Within the prefrontal association cortex, there is a topographic difference in activation pattern with the rostral areas (BA 9, 10) being more active than the caudal areas (BA 11, 12) over the relative time course. The temporal relationship within the limbic association cortex itself can also be observed using FIDSP to show that the anterior cinglate gyrus is activated earlier than the medial and posterior cingulate gyrus .

[0061] Due to the temporal aspect of activation, non- temporal analysis (e.g., conventional time-course driven analysis) cannot consistently extract the relevant ROA related to pain. In other words, a snapshot image obtained from conventional experimental time-course driven analysis tends to underestimate the number of ROAs relevant in pain. In some implementations, using temporal dynamic activation analysis, according^' to the specification, these different degrees of activation at different relative time points can be consistently identified and determined. PIDSP enables analysis with detailed information regarding the temporal dynamic relationships between cortex of different hemispheres, cortex of different functionalities, and various BAs within the same cortex. These information can be crucial in understanding how peripheral thermal pain stimulus is being processed and modulated in the brain and how these relationship can change in chronic pain states. [0062] The results of FIDSP in performing temporal dynamic activation analysis can be presented graphically as shown in FIGS. 3-10. The total number of voxel clusters within each relevant BA at a relative time course is used for constructing various three dimensional surface contour plots for illustrating the temporal dynamic relationship between different BAs at 258.

[0063] In some implementations, graphic presentation of the result from FIDSP may include either a grouped or individual ROA picture with the highest time-course correlation, and a three-dimensional (3-D) surface contour plot to illustrate the temporal dynamic information for different ROAs.

Graphical Presentations of PIDSP Analysis

[0064] FIGS. 3 through 10 are graphically presented results of a FIDSP analysis performed on various functional image data (e.g., fMRI data.) The functional image data are obtained from a group of 10 subjects (5 males and 5 females) enrolled in various pain model studies. Due to the precision of FIDSP and the added information provided by using temporal dynamic processing, the resultant graphical presentation shows involvements (by showing specific activations) of the relevant ROAs in central pain processing. The median age for the cohort is 30 years old with a range of 18 to 45 years old. The baseline cold, warm, cold pain, and hot pain thresholds for each subject are measured along the left calf. [0065] FIG. 3 is an image of the brain depicting stimuli- related activation of relevant grouped ROAs (Brodmann's areas) with uncorrected P<0.001 and spatial extent >10 voxels. The highlighted (darker shadings) portions of the brain represent the relevant BAs . The BAs includes bilateral primary (BAs 1,2,3) and secondary somatosensory (BA 5) cortices, dorsal lateral prefrontal cortex (BA 46) , anterior and medial cingulate gyrus (BAs 23, 24, 32), amygadala, basal ganglion, premotor cortex (BA 6), limbic cortex (BAs 34,38) and thalamus. A robust response (e.g., activation) in bilateral visual cortices is also shown.

Graphical Representations of Temporal Dynamic Analysis of ROAs [0066] FIG. 4 is a template of Brodmann's Area (BA) with each BA labeled using a unique number (e.g., 1-7, 9-12, 17-19, 23-33, and 45-47) to identify the location of each BA. For the BAs shown in the template, three-dimensional surface contour plots are generated to illustrate the temporal dynamic information/relations as shown in Figures 5-10. In Figures 5- 10, the X-axis represents the relative time point; the Y-axis represents the number of voxel clusters; and the Z-axis represents different BAs (correlated to the BA identified in the title of each figure.)

[0067] FIG. 5 shows 3-D surface contour plots representing temporal dynamic responses constructed for BAs 1-9 located in the Somatosensory Cortex and Motor Cortex. The 3-D contour plots show early bilateral primary somatosensory cortex (BA 1, 2, 3) activation. Over the relative time course observed, the rostral primary SSC (BA 3) is shown to be more active and sustained in its activities than caudal SSC (BA 1, 2) . In the secondary and tertiary somatosensory cortex (BA 5, I)₁ bilateral early activation is shown with the right side being more sustained in overall activities. The contour plots show similar bilateral temporal activation patterns in the primary motor cortices. However in the premotor and supplemental motor cortices (BA 6) a robust multiphasic temporal activation pattern is shown on the left cortex, whereas, on the right side (BA 6) , a milder sustained activation pattern is generated after an initial spike.

[0068] FIG. 6 shows 3-D surface contour plots for the Prefrontal Association Cortex (BAs 9-12) . A biphasic activation pattern is identified in bilateral prefrontal cortex with the rostral areas (BAs 9, 10) being more active than the caudal areas (BAs 11, 12) . The onset of activation of the prefrontal cortex is shown to be later than primary somatosensory and motor cortices.

[0069] FIG. 7 shows 3-D surface contour plots for the Limbic Association cortex (BAs 23-33) . Bilateral limbic association cortex activations are shown in the plots. The plots show on the left side, an early activation in the anterior cingulate gyrus (BA 32) , which is followed by an increase of voxel activations in the posterior cingulate gyrus

(BA 31) at the end of the relative time course. The plots also show that the peak activation of the anterior cingulate gyrus on the right side has a slight delay in comparison to left (BA 32) . Further the plots show that the middle and posterior have a multiphasic and sustained activation pattern throughout the relative time course.

[0070] FIG. 8 shows 3-D surface contour plots of the prefrontal cortex (BAs 45-46) . With thermal pain stimulations, the activation patterns of bilateral dorsolateral prefrontal cortices (BA 46) activation are highly similar as shown in FIG. 8. In addition, the plots show a biphasic activation pattern in the bilateral prefrontal association cortices with initial peak at the second relative time point.

[0071] FIG. 9 illustrates 3-D surface contour plots of a bilateral cerebellum with similar dynamic pattern of activation. Sustained patterns of activation noted in bilateral lingual, tuber, tonsil, culmen and uvula area of the cerebellum are also shown.

[0072] FIG. 10 shows 3-D surface contour plots illustrating and interpreting the visual cortex (BAs 9-12) . The contour plots show a presence of a sustained and late onset of bilateral visual cortex.

Increased Sensitivity and Specificity of FIDSP [0073] FIDSP provides greater sensitivity and specificity in identifying the relevant ROAs in central pain processing (see Figures 3-10) than conventional processing methods. In particular, FIDSP implements fMRILab, to identify the known ROAs involved in thermal pain (e.g., bilateral primary and secondary somatosensory cortex, motor cortex, anterior cingulate gyrus and thalamus) together in correspondence to the subject button-press time course. Because conventional processing methods typically are not able to identify all of the relevant ROAs together, FIDSP provides a greater degree of sensitive and specificity in identifying the relevant ROAs related to pain.

[0074] In addition, when ROAs are constructed based on a group comparison (e.g., of ten subjects), results of FIDSP analysis show that the individual ROAs related to pain are quite consistent across the ten subjects. This observation further suggests that the specificity of FIDSP procedure of the identified pain related ROA is very high. [0075] Also, the increased specificity and sensitivity of FIDSP is further demonstrated by the ability of FIDSP to provide information not typically available using conventional analytical processing methods. For example, the results of FIDSP can be analyzed to show that the ROAs related to thermal pain stimuli tend to be bilateral with more robust activation on the contralateral (right) side of the cortex to the stimulus (left calf) .

[0076] Further, FIDSP can be implemented to identify previously unidentified ROAs relevant to central pain processing. For example, FIDSP can be implemented to determine the role of prefrontal cortex and cerebellum in pain processing and modulation. The visual cortex, in particular, are very active for all subjects under thermal stimulation even though all subjects are covered with an eye-shield during the study. This may be due to the fact that the subjects, although being blind-folded with an eye shield, were trying to visualize the source of noxious stimulation without any physical visual stimulation. Alternatively, the visual cortex plays an important role in pain modulation.

FIDSP Usage

[0077] FIG. HA is a flowchart of a process 1100 for implementing FIDSP according to the specification. A user takes functional imaging data (e.g., fMRI data) and separates the data into independent components using ICA decomposition at 1110. The decomposed components are analyzed by creating log files for FIDSP data analysis at 1120. The created log files are used to execute FIDSP data analysis at 1140. The results of the FIDSP data analysis are saved as multiple summary text files. The saved text files are transferred to a spreadsheet to sort and organize the FIDSP analysis data at 1170. For example, the saved text files can be transferred to a Windows'⁸ computer and pasted into an Excel^® file for sorting the FIDSP analysis data. The generated Excel^® file is used to perform a second level analysis at 1190. [0078] FIG. HB is a flowchart for performing ICA decomposition 1110. Functional imaging data is received using fMRLAB at 1112. The type of functional imaging data (e.g., fMRI, functional PET, etc.) is identified at 1114. The computer system 110 can be implemented to automatically identify the functional imaging type by using fMRLAB or other executable program working in conjunction with fMRLAB. The identified functional imaging data (e.g., raw fMRI data) is decomposed into independent components using fMRLAB under Matlab at 1116. A total of 150 to 200 components can be extracted from the functional imaging data on average. All components are saved under a XXXX. fMRI file for post-ICA temporal dynamic construction at 1118.

[0079] FIG. HC is a flowchart for performing post-ICA decomposition and creating log (e.g., text) files for FIDSP analysis 1120. The user opens the post-ICA decomposition DATA folder and searches for the specific subject and study session to be analyzed at 1122. The user opens a folder named 'LOGS' at 1124 and searches for the desired log file to open. Each log file is named by the study to which it belongs, such as "acup.log" or "acup_thermol.log." (See FIG. 12 for sample log folder.) Once the log file is active, a graphical user interface (GUI) , such as a pop-up window is displayed to present the user with an option of either executing the log file or displaying its contents at 1126. Selecting the DISPLAY option opens the log file in a text editor. (See FIG. 13 for a sample log file.)

[0080] Using the text editor, the log file can be modified by the user at 1128 as needed. For example, all extraneous information are deleted to preserve only the substantive information in the proper format. The user may delete all the information (horizontally) which is not labeled as "RESPONSE' under the "Event" column.- The user may also delete all the information (vertically) which is not listed under the "Time" column. Deleting the extraneous information leaves the log file having an even number of time points . For formatting purpose, the time points are modified to delete the last digit of each time point. The modified time points are in milliseconds (ms) . From each time point, 7000 ms is subtracted to normalize the time points. The modified Log (text) file is saved at 1130 with a clearly labeled name, such as "atl.txt" for acupuncture-thermo 1 analysis. This modified log file is now complete and can be used in FIDSP data analysis. (See FIG. 14 for a sample log.txt file.) [0081] FIG. HD is a flowchart for executing FIDSP data analysis 1140. FIDSP analysis is executed under Linux in the following example, but other suitable operating systems are within the scope of the specification. A terminal window is opened by the user by right-clicking on the desktop and selecting "Open Terminal Window" at 1142. The MATLAB^® program is executed by entering 'MATLAB' into the terminal window and pressing enter at 1144. At this point one terminal window and one MATLAB^* command window are open. While the user will no long user interact with the terminal window, the terminal window must stay open since MATLAB is executing in connection with the terminal window and closing the terminal window will also terminate MATLAB*. (See FIG. 15 for a sample screen.) [0082] Before performing FIDSP data processing, the user verifies that FIDSP directory, which contains all the functions for data analysis, is designated as the "current directory' at 1146. From the FIDSP directory one or more desired functions are called or accessed at 1148. In particular, the "acup_thermo_postica ()" function can be accessed by entering various MATLAB argument commands (e.g., a set of three arguments) . For example, when analyzing acupuncture for subject "0505081k¹ s" folder, which is located in the data folder, the following argument command is entered into MATLAB^® command window.

"acup_thermo_j)ostica( ' data/050508lk/acup. f mr¹ , 'data/050508lk/logs/a.txt • ,0:1000:1500 0) ;^»

[0083] If for example, the analysis involves thertno or acupuncture-thermo related data, the following argument command is entered into the MATLAB command window.

"acup_thermo_postica ( ' data/0505081k/acup . f mr' , 'data/050508lk/logs/a.txt ' ) "

[0084] It should be noted, however, that these are not constant arguments and the names of the folders which contain the data to be analyzed must be entered. (See FIG. 16 for an example . )

[0085] Once the appropriate argument command has been entered, a GUI box (e.g., a pop up window) is presented to the user at 1150 with a graphical representation (e.g., an image, a picture, etc.) of the brain and various GUI input elements (e.g., buttons) for moving the image of the brain about X, Y, or Z axis. By interacting with appropriate buttons, the image of the brain is rotated around the correct axis to align the image with the front side/face pointing toward the user at 1152. In most cases, two rotations about the z-axis positions the image as desired. Once the image is properly rotated and aligned, a button labeled "flip about Y" followed by a button labeled "OK" are selected in sequence to signal the MATLAB^® program to write relevant ROA files at 1154. (See FIG. 17 for a sample screen.)

[0086] Without closing the MATLAB^® command window or the terminal window used to start MATLAB^* , a second terminal window is opened at 1156. The "current directory" is verified again to make sure the correct directory type (FIDSP directory) has been specified, and then a series of commands are entered to execute writing of all ROA images at 1158. At the command line, the "tcsh" command is entered (including the "enter" key press) . When a new command line appears, the "source fsi . csh" command is entered in the new command line. When the next new command line appears, the command, " . /batch_sm. csh" is entered. In response to the entered commands, a message that states "estimating spatial normalization parameters" is displayed to the user. This message remains displayed until all of the ROA images are written and a new command line appears. (See FIG. 18 for a sample end screen.) [0087] The user returns to the MATLAB command window to finish the analysis and generate output data as nroa . txt files at 1160. The "acup_thermo__brainareas () " function is executed by entering the following argument command,

"acup_thermo_brainareas (1.5)" in the MATLAB command line. Once all of the nroa. text files have been written and a new command line appears, the "acup_thermo_sumup" function is executed by entering "acup_thermo_sumup" in the new command line. FIDSP analysis is now complete, and the FIDSP directory contains the desired output information in the corresponding text files. From the multiple text files generated in the FIDSP directory, three text files, "summary.txt", "summ_neg.txt", and "summ_pos.txt", are saved for further processing and output generation at 1162. (See FIG. 19 for sample final screen in MATLAB . )

[0088] FIG. HE is a flow chart for transferring FIDSP analysis data (e.g., the three saved text files) 1170. The contents of the saved text files are transferring to a spreadsheet (e.g., Excel^" program executing on a Windows computer.) to sort and organize data. The saved files are opened using Excel^® or other suitable spreadsheet program at 1172. If needed, the saved text files can be transferred to a different computer executing Excel⁸. Transferring the text files can be accomplished by copying the text files onto a computer readable media (e.g., a disk, CD, portable flash memory device, or other data storage devices) , and reading the stored text files on a different computing device. Alternatively, the saved text files can be transferred to a different computing device using wired or wireless data transfer (e.g., using local network, WiFi, Bluetooth, etc.) [0089] The contents of the saved text files are sorted using the sort function in Excel at 1174. The data in each row of the text files opened in Excel contain a time component in parenthesis. In the cell above each time component, a label from 1 to 5 going from the smallest time value to the largest is applied. Using the sort option, "sorts left to right" is executed to sort the data in the saved text files. The result is one or more readable and organized excel files from which data can be extracted.

[0090] FIG. HF is a flowchart for performing second level

® analysis 1190 on the sorted text files. Under Excel , number of voxels, either positive correlated voxels (PCV) or negative correlated voxels (NCV) in the regions of activities (ROAs) are grouped under their corresponding Brodmann area (BA) at 1192. A ratio of (number of NCV) /number of (NCV+PCV) is calculated using Excel^® calculated at 1194. At 1196, a 2- factor repeat measures of ANOVA is applied to assess the significant interaction of time (relative time point) and side (left vs. right) . Significant main effects and interactions were further analyzed, at 1198, by using appropriate tests of simple main effects and/or interaction contrasts as dictated by the outcome of the overall ANOVA. A two-tailed p-value of 0.05 was considered significant.

[0091] One or more of the following graphs: 1) PCV, 2) NCV, 3) PCV+NCV, or 4) PCV+NCV+Ratio over the relative time point can be generated and displayed at 1199 to indicate the dynamic relationship of each ROA at each relative time point. The graphs of different ROAs can be used for temporal dynamic comparison.

Applications of FIDSP

[0092] Using ICA and behavioral feedback measurement, FIDSP can be implemented to successfully identify the ROA related to pain with high sensitivity and specificity. The established temporal dynamic activation pattern of ROA related to pain can be used as a baseline data for studying how these dynamic patterns changed in chronic pain states and how the processing pattern changes in response to non-pharmacological or pharmacological interventions. Moreover, the FIDSP can also be used for studying temporal dynamic of other modalities of cognitive sciences or diseases states in which central mechanisms may play an important role.

[0093] FIDSP may be implemented to process other functional neuroimaging data (other than fMRI) including functional PET, functional SPECT (single photon emission computer tomography) , and functional NIR (near infrared) imaging. Further FIDSP may be applied to a wide range of applications including (1) facilitating drug screening in both animal and human pain models (e.g., drug screening between phase I and phase II of a clinical trial as a small pilot study to determine what is promising and most relevant to the processing and perception of pain) ; (2) on-site data analysis of fMRI scanners; (3) a diagnostic biomarker for chronic pain states; (4) integration into MRI scanners to provide additional diagnostic functions for physicians without specialized training in fMRI; and (5) evaluation of behavioral pain scores and with most relevant Brodmann regions of the brain.

[0094] In terms of clinical application, FIDSP can potentially establish specific pain processing dynamic pattern for certain chronic pain conditions such as certain common neuropathic pain states (e.g. complex regional pain syndrome, post-stroke central pain syndrome, phantom limb pain, postherpetic neuralgia, etc.), mechanically induced pain (osteoarthritis) , and inflammatory pain (post-surgical pain or post-injury pain) . Identifying the difference in pain processing among different chronic pain conditions can lead to better understanding of chronic pain mechanisms and allow development of mechanism-specific analgesics. [0095] Various implementations of the subject matter described herein may be realized in digital electronic circuitry, integrated circuitry, specially designed ASICs (application specific integrated circuits) , computer hardware, firmware, software, and/or combinations thereof. These various implementations may include implementation in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device. [0096] These computer programs (also known as programs, software, software applications, or code) include machine instructions for a programmable processor, and may be implemented in a high-level procedural and/or object-oriented programming language, and/or in assembly/machine language. As used herein, the term "information carrier" comprises a "machine-readable medium" that includes any computer program product, apparatus and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal, as well as a propagated machine-readable signal. The term "machine- readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor. [0097] To provide for interaction with a user, the subject matter described herein may be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user may provide input to the computer. Other kinds of devices may be used to provide for interaction with a user as well; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback) ; and input from the user may be received in any form, including acoustic, speech, or tactile input. [0098] The subject matter described herein may be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a client computer having a graphical user interface or a Web browser through which a user may interact with an implementation of the subject matter described herein), or any combination of such back-end, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a WAN, and the Internet. [0099] The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. [00100] While this specification contains many specifics, these should not be construed as limitations on the scope of any described technique or of what may be claimed, but rather as descriptions of features specific to particular embodiments. Certain features that are described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a subcombination or variation of a subcombination.

[00101] Similarly, while operations are depicted in the drawings in a particular order, this should not be understand as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[00102] In addition to these variations, other modifications are possible and within the scope of the following claims.

Claims

ClaimsClaims

1 . A method comprising : receiving functional imaging data; separating the received imaging data into a plurality of components; selecting one or more of the components; spatially normalizing the selected components; constructing grouped regions of activity (ROAs) based on a selected spatially normalized component; determining one or more Brodmann areas (BAs) in the constructed grouped ROAs; and constructing temporal activities of the determined BAs.

2. The method of claim 1, wherein receiving the functional imaging data comprises receiving data from a functional magnetic imaging (fMRI) system.

3. The method of claim 1, wherein receiving the functional imaging data comprises receiving data from a non-MRI system.

4. The method of claim 3, wherein receiving the functional imaging data from a non-MRI system comprises receiving data from a functional positron emission tomography (PET) system.

5. The method of claim 3, wherein receiving the functional imaging data from a non-MRI system comprises receiving data from a functional single photon emission computer tomography (SPECT) system.

6. The method of claim 3, wherein receiving the functional imaging data from a non-MRI system comprises receiving data from a functional near infrared imaging (NIR) system.

7. The method of claim 1, wherein separating the received functional imaging data further comprises separating the received functional imaging data into a plurality of components, each component having a unique time course and a region of activity (ROA) .

8. The method of claim 7, wherein separating the received functional imaging data comprises performing independent component analysis (ICA) .

9. The method of claim 7, wherein receiving the functional imaging data further comprises : measuring one or more baseline sensory thresholds of a human subj ect ; and capturing one or more functional images of the human subject while applying stimuli to the human subject, wherein the applied stimuli are based on the measured baseline sensory thresholds .

10. The method of claim 9, wherein capturing the functional images further comprises determining a time course of the human subject indicating an onset of pain.

11. The method of claim 10, wherein selecting one or more of the components comprises: correlating the time course of each component with the time course of the human subject indicating an onset and offset of pain to determine a linear correlation coefficient for each component; determining a preset threshold of the correlation coefficient; and selecting the one or more components based on the preset threshold.

12. The method of claim 11, wherein constructing grouped ROAs comprises selecting a spatially normalized component having a highest level of linear correlation.

13. The method of claim 1, further comprising determining a total number of voxel clusters present in each BA; and generating an output based on the determined total number of voxel clusters .

14. The method of claim 13, wherein generating an output comprises constructing one or more three-dimensional surface contour plots.

15. A computer program product, encoded on a computer- readable medium, operable to cause data processing apparatus to perform operations comprising: receiving functional imaging data; separating the received functional imaging data into a plurality of components; selecting one or more of the components based on a threshold; spatially normalizing the selected components; constructing grouped regions of activity (ROAs) based on a selected spatially normalized component; determining one or more Brodmann areas (BAs) in the constructed grouped ROAs; and constructing temporal activities of the determined BAs.

16. The computer program product of claim 15, wherein the product is further operable to cause data processing apparatus to perform operations comprising receiving data from a functional magnetic resonance imaging (fMRI) system.

17. The computer program product of claim 15, wherein the product is further operable to cause data processing apparatus to perform operations comprising receiving data from a non-MRI imaging system.

18. The computer program product of claim 17, wherein the product is further operable to cause data processing apparatus to perform operations comprising receiving data from a functional positron emission tomography (PET) system.

19. The computer program product of claim 17, wherein the product is further operable to cause data processing apparatus to perform operations comprising receiving data from a functional single photon emission computer tomography (SPECT) system.

20. The computer program product o claim 17, wherein the product is further operable to cause data processing apparatus to perform operations comprising receiving data from a functional near infrared imaging (NIR) system.

21. The computer program product of claim 15, wherein the product is further operable to cause data processing apparatus to perform operations comprising separating the received functional imaging data to obtain one or more components, each having a unique time course and a region of activity (ROA) .

22. The computer program product of claim 21, wherein the product is further operable to cause data processing apparatus to perform operations comprising separating the received functional imaging data into a plurality of components using independent component analysis (ICA) .

23. The computer program product of claim 21, wherein the product is further operable to cause data processing apparatus to perform operations comprising: measuring one or more baseline sensory thresholds of a human subject; and capturing one or more functional images of the human subject while applying stimuli to the human subject, wherein the applied stimuli are based on the measured baseline sensory thresholds .

24. The computer program product of claim 23, wherein the product is further operable to cause data processing apparatus to perform operations comprising determining a time course of the human subject indicating an onset and offset of pain.

25. The computer program product of claim 24, wherein the product is further operable to cause data processing apparatus to perform operations comprising: correlating the time course of each component with the time course of the human subject indicating an onset and offset of pain to determine a linear correlation coefficient for each component ; determining a preset threshold of the correlation coefficient; and selecting the one or more components based on the preset threshold.

26. The computer program product of claim 25, wherein the product is further operable to cause data processing apparatus to perform operations comprising selecting a spatially normalized component having highest level of linear correlation coefficient.

27. The computer program product of claim 15, wherein the product is further operable to cause data processing apparatus to perform operations comprising determining a total number of voxel clusters present in each BA; and generating an output based on the determined total number of voxel clusters .

28. The computer program product of claim 27, wherein the product is further operable to cause data processing apparatus to perform operations comprising creating one or more three-dimensional surface contour plots.

29. A system comprising: a user interface device; and one or more computers communicatively coupled to the user interface device, the one or more computers including a processor configured to process functional imaging data, wherein processing the functional imaging data includes receiving functional imaging data; separating the received functional imaging data into a plurality of components; selecting one or more of the components; spatially normalizing the selected components; constructing grouped regions of activity (ROAs) based on a selected spatially normalized component; determining one or more Brodmann areas (BAs) in the constructed grouped ROAs; and constructing temporal activities of the determined BAs.

30. The system of claim 29, wherein processing the functional imaging data further includes determining a total number of voxel clusters present in each BA; and constructing one or more three-dimensional surface contour plots based on the determined total number of voxel clusters .

31. The system of claim 29, wherein the functional imaging data is received from a functional magnetic resonance imaging (fMRI) system.

32. The system of claim 29, wherein the functional imaging data is received from a non-MRI system.

33. The system of claim 32, wherein a non-MRI system comprises a functional positron emission tomography (PET) system.

34. The system of claim 32, wherein a non-MRI system comprises a functional single photon emission computer tomography (SPECT) system.

35. The system of claim 33, wherein a non-MRI system comprises a functional near infrared imaging (NIR) system.