Autocalibrating multi-echo functional magnetic resonance imaging

Abstract

Methods and systems for auto-calibrated functional magnetic resonance imaging. Combined time series data is computed using a weighted sum of the values of each of the voxels across a plurality of echo times in volumetric time series data describing the blood-oxygen-level-dependent (BOLD) values of each of a plurality of voxels in a brain at a plurality of echo times. The combined time series data is decomposed using independent component analysis to identify neural signals and the identified neural signals is converted into units of / ?2 and provided as output to an operator.

Description

AUTOCALIBRATING MULTI-ECHO FUNCTIONAL MAGNETIC RESONANCE IMAGINGCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] The present application claims the benefit of and priority to United States Provisional Application No. 63 / 599,510, filed on November 15, 2023, and United States Provisional Application No. 63 / 554,964, filed on February 17, 2024, the entire disclosure of each of which is hereby incorporated by reference as if set forth in its entirety herein.TECHNICAL FIELD

[0002] Embodiments described herein generally relate to methods and apparatus for quantitative functional magnetic resonance imaging (fMRI), and more specifically to the methods of quantitative fMRI which allow comparability of measurements across different imaging systems and scanners.BACKGROUND

[0003] Functional magnetic resonance imaging (fMRI) scans the brain in slices, repeatedly, to reconstruct 3 -dimensional volumes of the brain sequentially to track changes in brain activity over time. For each repetition, a radiofrequency pulse is applied to excite proton spins in neural tissue, and images are acquired to capture signal across brain tissue as the signal decays. However, nonideal conditions in brain anatomy, patient motion, scanner errors, and pulse sequence design disrupt T2* neural) signals and produce non-T2(i e., non-neural) signals that often dominate datasets.

[0004] To mitigate the impact of these issues on the resulting data, fMRI post-acquisition processing can include a variety of techniques including spatial smoothing, fitting noise models, and applying frequency-domain filters. The rationale underlying processing choices is imprecise and highly variable, relying on indirect fixes that produce unreliable results. Analyzing conventional fMRI data through eliminative filtering provides an indirect measurement of the effect of interest which must be reproduced over a long acquisition among many subjects to confirm its validity.

[0005] Ordinarily quantification with fMRI is not feasible due its inability to determine true blood oxygen level dependent (BOLD) origin of signals without references, whether anatomical or temporal. The situation is further complicated by the dominance of non-BOLD signals and theirneed to be removed. For these reasons, conventional fMRI sequences are limited in their ability to provide quantitatively meaningful metrics of brain activity.

[0006] Accordingly, a need exists for improved methods and systems for fMRI that improve signal quality and reduce noise and artifacts.SUMMARY

[0007] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description section. This summary is not intended to identify or exclude key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0008] The methods of multi-echo functional MRI (ME-fMRI) and ME-ICA (an open source pipeline that handles the preprocessing and analysis of ME-fMRI data available for download at: https: / / me-ica.readthedocs.io / en / latest / ) provide fMRI time series with very low artifact content and high signal-to-noise ratio based on an effective isolation of BOLD signals based on a physiological model. These models are quantitative and can serve as a basis for NMR relaxometry, which is a solid-state NMR method of characterizing sample properties based on NMR relaxation rates, which can also be measured by MRI in appropriately designed experiments

[0009] Embodiments of the present invention provide strategies to quantify fMRI signals to a physiological basis related to the relaxation rate due to metabolic activity affecting tissue iron, given only small changes in signal of interest that are predominantly driven by neurovascular response to neural activity. This involves treatments of data from gradient-echo EPI.

[0010] This approach enables a variety of applications, such as fMRI studies for: a single subject at different times or before and after intervention; for multiple subjects at the same site or same time or before and after intervention; for different subjects at different sites using different scanners; studies comparing individuals to groups; and longitudinal studies, such as those monitoring disease progression.

[0011] In one aspect, embodiments relate to a method for auto-calibrated functional magnetic resonance imaging. The method includes receiving volumetric time series data describing the BOLD values of each of a plurality of voxels in a brain at a plurality of echo times; computing a combined time series data using a weighted sum of the values of each of the voxels across the plurality of echo times; decomposing the combined time series data using independent component analysis to identify neural signals; converting the combined time series data into units of R2, and providing an output incorporating the identified neural signals.

[0012] In some embodiments the volumetric time series data is received from a magnetic resonance imaging scanner or a data store.

[0013] In some embodiments the method includes preprocessing the volumetric time series data. In some embodiments preprocessing includes at least one of rigid body motion correction, T2* -driven functional-anatomical co-regi strati on, de-obliquing, or slice-time correction.

[0014] In some embodiments the coefficients of the weighted sum arew^ TE where TEtis an echo time of interest, n is the number of theplurality of echo times, and T2is the susceptibility-weighted transverse relaxation contrast.

[0015] In some embodiments the method includes applying principal components analysis to the combined time series data to remove noise.

[0016] In some embodiments the method includes identifying neural signals using ICA and TE-dependence metrics to identify components with significant BOLD weighting and nonsignificant non-BOLD weighting.

[0017] In some embodiments providing an output includes generating reassembled volumetric time series data from components with significant BOLD weighting and non-significant non- BOLD weighting while discarding components with non-significant BOLD weighting and with or without high non-BOLD weighting. In some embodiments the output is at least one of an anatomical MRI image, a spatial map of an identified neural signal, a plurality of denoised time series images, or a report summarizing data quality and network activity.

[0018] In some embodiments the volumetric time series data includes at least one of data from a single subject taken at different times; data from a plurality of subjects taken at the same time or the same site; unprocessed data from a plurality of scanners, wherein at least two of the plurality are different models of scanner.

[0019] In another aspect, embodiments relate to a system for auto-calibrated functional magnetic resonance imaging. The system includes a processor; a user interface; and a memory storing instructions that, when executed by the processor, cause the processor to receive volumetric time series data describing the BOLD values of each of a plurality of voxels in a brain at a plurality of echo times; compute a combined time series data using a weighted sum of the values of each of the voxels across the plurality of echo times; decompose the combined time series data using independent component analysis to identify neural signals; convert the combined time series data into units of / ?2, and provide an output incorporating the identified neural signals.

[0020] In some embodiments the system further includes a network interface and the volumetric time series data is received from a magnetic resonance imaging scanner or a data store via the network interface.

[0021] In some embodiments the memory further includes instructions that, when executed by the processor, cause the processor to preprocess the volumetric time series data. In some embodiments preprocessing includes at least one of include rigid body motion correction, T2- driven functional-anatomical co-regi strati on, de-obliquing, or slice-time correction.

[0022] In some embodiments the coefficients of the weighted sum are^, TEdwhere TE. is an echo time of interest, n is the number of theplurality of echo times, and T2is the susceptibility-weighted transverse relaxation contrast.

[0023] In some embodiments the memory further includes instructions that, when executed by the processor, cause the processor to apply principal components analysis to the combined time series data to remove noise.

[0024] In some embodiments the memory further includes instructions that, when executed by the processor, cause the processor to identify neural signals using ICA and TE-dependence metrics to identify components with significant BOLD weighting and non-significant non-BOLD weighting.

[0025] In some embodiments providing an output includes generating reassembled volumetric time series data from components with significant BOLD weighting and non-significant non- BOLD weighting while discarding components with non-significant BOLD weighting and with or without high non-BOLD weighting. In some embodiments the output is at least one of an anatomical MRI image, a spatial map of an identified neural signal, a plurality of denoised time series images, or a report summarizing data quality and network activity.

[0026] In some embodiments the volumetric time series data includes at least one of: data from a single subject taken at different times; data from a plurality of subjects taken at the same time or the same site; unprocessed data from a plurality of scanners, wherein at least two of the plurality are different models of scanner.BRIEF DESCRIPTION OF DRAWINGS

[0027] Non-limiting and non-exhaustive embodiments of this disclosure are described with reference to the following figures, wherein like reference numerals refer to like parts throughout the various views unless otherwise specified:

[0028] Figure l is a flowchart of a method in accord with the present invention;

[0029] Figure 2 is a block diagram of an exemplary apparatus; and

[0030] Figure 3 depicts the apparatus of Figure 2 deployed in an exemplary hospital environment.DETAILED DESCRIPTION

[0031] Various embodiments are described more fully below with reference to the accompanying drawings, which form a part hereof, and which show specific exemplary embodiments. However, the concepts of the present disclosure may be implemented in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided as part of a thorough and complete disclosure, to fully convey the scope of the concepts, techniques and implementations of the present disclosure to those skilled in the art. Embodiments may be practiced as methods, systems or devices. The following detailed description is, therefore, not to be taken in a limiting sense.

[0032] Reference in the specification to “one embodiment” or to “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiments is included in at least one example implementation or technique in accordance with the present disclosure. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment.

[0033] In addition, the language used in the specification has been principally selected for readability and instructional purposes and may not have been selected to delineate or circumscribe the disclosed subject matter. Accordingly, the present disclosure is intended to be illustrative, and not limiting, of the scope of the concepts discussed herein.Overview

[0034] Embodiments of the present invention use multi-echo (ME) fMRI, a variation of more typical single-echo fMRI data acquisition, to improve the feasibility and reliability of fMRI exams chiefly through mitigation of a wide variety of fMRI imaging artifacts that reduce the interpretability and thus efficacy of fMRI in clinical practice. In ME-fMRI, three or more images are acquired at multiple echo times (TE) in the same time window to capture signal across the decay period. In contrast, standard single-echo fMRI images are acquired at one TE.

[0035] BOLD signals have a characteristic relationship with TE, while non-BOLD noise signals do not. Embodiments of the present invention use the additional information from ME-fMRI to systematically disentangle the neuronal -related signals (T2) from non-neuronal artifactual signals (non-T2) based on the biophysical model of signal decay itself without user intervention and additional denoising steps. These techniques not only improve the identification of true neural signal, but also allow for quantitative measurement of the change in blood deoxygenation state in radiological units of magnetic susceptibility ( / ?2) to provide absolute measurements of stable signal in quantitatively meaningful units linked to perfusion dynamics that can be tracked over time, as affected by tasks, between different brain states, between patients or groups of patients, across sites, across different scanners, etc.

[0036] The result is a measurement of brain activity dynamics presented in voxel time series based on precise measurements of the BOLD effect with high spatial and temporal resolution in terms of / ?2and time-varying fluctuations quantified as rate changes A / ?2, both in units of milliseconds.

[0037] Time series data produced by embodiments of the present invention can be used for task-based and resting-state analyses with the advantages of artifact-free signal, major increases in contrast-to-noise, and quantitative reliability. This data can also be used to produce reports that summarize the data quality and provide a whole-brain functional network quantitation, tabulating the activity magnitudes in a set of canonical brain networks to summarize the extent of each network’s expression in the individual. This summary can be used to determine whether a brain network expresses a magnitude that is typical as compared to a large sample of healthy adults for a known condition or to track the magnitude of a given network across scans.Embodiments

[0038] Figure 1 is a flowchart of an exemplary method 100 for autocalibrating fMRI. The process begins with the receipt of BOLD time series data at multiple echo times in the same run, with each voxel having multiple time series, one per echo time (Step 104).

[0039] In one embodiment, the BOLD time series data can be acquired in an active hospital environment using MRI equipment. The requisite MRI exam is non-contrast and like all MRI is non-ionizing using radiofrequency (RF) signals, compatible with standard clinical 3T MRI systems with no specialized hardware, requiring instead the use of specific pulse sequences installed and / or enabled on the MRI scanner.

[0040] The multi-echo fMRI (ME-fMRI) acquires data for BOLD signals using susceptibility- weighted transverse relaxation (T2) contrast with echo planar imaging (EPI), where EPI is acquired repeatedly after each acquisition such that BOLD signals across multiple contrasts are acquiredper data frame. By contrast, standard fMRI acquires one contrast per image, often wasting wait time after initial excitation to do so, thereby losing information and necessitating the use of methods such as those described herein to remove imaging artifacts.

[0041] In yet another embodiment the BOLD time series data has been previously acquired using the aforementioned ME-fMRI scans and stored in persistent storage such as NAS or cloud storage. The scans may be stored using standard Radiology Information System (RIS) procedure in DICOM or another form.

[0042] The BOLD time series data taken at the plurality of TEs is then combined to produce a combined image time series (Step 108) that retains signal from the vast majority of the brain including brain regions with signals lost in conventional fMRI.

[0043] Initial preprocessing of the volumetric timeseries data is relatively minimal as downstream signal denoising performs best when the fMRI data remain as close to their acquisition form as possible. Preprocessing adapts basic (i.e., requisite) steps to keep the timeseries from all TEs in alignment prior to combination and denoising. These steps include rigid body motion correction, T2-driven functional-anatomical co-registration, de-obliquing, and slice-time correction. Additional customary processing steps such as spatial smoothing, temporal detrending, and bandpass filtering are not performed as they are inherently managed in the process of signal denoising.

[0044] The relative susceptibility of a voxel to dephase at a given T Etversus at other TEs up to n can be expressed as a weight wwhich is used to compute a weighted sum of signals across TEs. The resulting summed signal not only emphasizes the best contributing signal at TE but over the volume approximates an acquisition where each voxel is acquired at the TE that is ideal for its T2.

[0045] The combined data is then denoised using principal components analysis (PCA) to remove noise signals (Step 112). PCA separates components as orthogonal, unit-normed, normally distributed correlated sources. In practice, PCA identifies high variance components that reflect mixtures of BOLD and non-BOLD physical signal processes and low variance components that are typically Gaussian distributed and reflect thermal or electrical noise. The high variance components are subjected to further analysis while the low variance components are discarded.

[0046] Given a multivariate time series model including both BOLD and non-BOLD components, these components can be differentiated through an extension of the TE-dependence signal models. For a given component c, F-statistic maps can be averaged with weights corresponding to component amplitudes, Bc 0C, z-transformed to Zc, evaluated for each voxel v and raised to a power (typically 2):

[0047] These summary statistics scale like their underlying F-statistics and in a single numerical value express the relative BOLD or non-BOLD weighting of a component.

[0048] Neural signals are identified in the denoised data set using the computed K and p values (Step 116). The identification process begins with independent component analysis (ICA) with the objective of neural signal isolation. ICA can be thought of as a higher order PCA where the sources of components are orthogonal but also approximately statistically independent. Ordinarily, statistical independence of signals is challenging to establish. In certain data types like fMRI, biological inference indicates that distinctly meaningful sources can be elucidated using ICA.

[0049] Sparsity is the distinguishing property of natural signals, which are smooth and correlated and therefore not random. ICA can be viewed as a lossless rotation of the orthonormal linear system from PCA that maximizes the difference in sparsity from the principal eigenvectors. The overall rotation is obtained in small rotations from the weights of projection to an incrementally sparser system by following the increasing gradient of the sparsity operator, i.e. as a fixed-point algorithm. ICA of PCA dimensionally-reduced data converge quickly following smooth curves.

[0050] The identified components fall into three regimes. The first component regime has significant BOLD weighting and non-significant non-BOLD weighting (K > ppf F, 0.05) A p S? ppf(F, 0.05)) and corresponds to BOLD signals. The second component regime has nonsignificant BOLD weighting regardless of the non-BOLD weighting (K < ppf(F, 0.05) V p > ppf(F, 0.05)) and corresponds to non-BOLD artifacts. The final component regime is a small transition band between the first two regimes (K ~ ppf(F, 0.05) Ap ~ ppf(F, 0.05)) , having some components associated with outlier levels of variance such as encoding errors.

[0051] To retain data with “pure” BOLD signal, the number of components is restricted to the first regime. Time series are reassembled with only these components for denoised images. Spatial maps for each BOLD component and the denoised time series are rendered as part of the processing outputs.

[0052] The component maps and time series data are then converted to R2 (Step 120), a radiological standard. fMRI activity is measured in units of percent signal change, which has an important biophysical basis but is highly influenced by parameter settings (e.g., TE) and interscanner variability. ME-fMRI data can be quantified in physical units of R2 (relaxation time) for baseline and A / ?2 for fluctuations from baseline.

[0053] The dominant signal in BOLD fMRI is known to be related to local (i.e., pixel-level) deoxyhemoglobin concentration, which means fMRI units in A / ?2 track changes in deoxyhemoglobin concentration due to perfusion, or the influx of blood water and efflux (washout) of deoxyhemoglobin. On this basis, a factor of ME-ICA relaxometry can be computed that expresses activation magnitude in terms of a change in deoxyhemoglobin concentration that is proportional to cerebral metabolism.

[0054] The equation of observed susceptibility-weighted transverse relaxation rate R2 to the concentration of iron is:where R2is the inverse of transverse relaxation time T2and <pFe* is the relaxivity of ferrous iron (II) in deoxyhemoglobin. Changes in concentration in Fe are due to perfusion, due to influx of blood water and efflux (washout) of deoxyhemoglobin.

[0055] BOLD signals increase related to the relative concentration of the diamagnetic species Hb“and ferrous iron, which is the product of metabolism. <pFe* is an experimental parameter that varies with scanner main field Bowhich is a function of pulse sequence. Therefore, when comparing datasets across the same pulse sequence and scanner, <pFe* can be assumed to be a linear constant, which means in linear treatments, e.g., least squares fitting for the general linear model, it does not need to be explicitly estimated. However, if comparing values from across different imaging conditions, e.g., scanners, pulse sequences, or sequence parameters then <pFe* needs to be estimated in order to stably estimate [Fe*].

[0056] The difference 4 / ? 2 related to a meaningful signal change can be represented in terms of quantification in Eq. 3:with the term R2 cancelling, which is relatively stable across tissue type but may be affected by partial-voluming. The change in iron concentration is then computed as:In practice, this calibrates for the impact of pulse sequence onand Boin determining AR2, to yield an estimate of standard activation A [Fe*] that happens to be in terms of changes in iron concentration.

[0057] If the effect of initial perfusion in a pixel is to be calibrated out to get a sense of “absolute” activation, a term for relative change in iron concentration would require rearranging Eq. 3 and an estimate R grayDefine R2' to be R2* — R2. The relaxivity term cancels giving:% 1[Fe*]act. = J7?5act / 7?2' [7] that depends only on assuming a value of ^2gray- Cancelling the relaxivity term accounts for any nonlinear effects of iron concentration across tissue states. The relaxation time T2of gray matter is notably stable across a wide range of condititions, including brain regions and even normal aging, but is dependent on field strength.

[0058] From Eq. 7, A [Fe*]actis sensitive to changes in neuronal activity of interest, age related variation of interest, systematic error (e.g., condition association) of [Fe*] in cerebrospinal fluid (CSF), and errors in masking of CSF and is robust to field strength, pulse sequence, and the effect of tissue type on measurement accuracy. This autocalibrated relaxometry unit (ARU) inherently calibrates brain activity across conditions, resulting in quantitatively meaningful units for comparison across equipment, timepoints, and even individuals. ARU boosts contrast-to-noise, differences across functional brain networks, and test-retest reliability.

[0059] Having successfully identified neural signals and converted them to ARUs, the results are suited to a variety of applications and may be supplied in a variety of formats (Step 124). Resulting outputs may include the initial anatomical MRI image, direct outputs of fMRI processing (i.e., timeseries), volumetric functional network maps, and summary reports.

[0060] A set of DICOM compliant denoised time series images may be output to accommodate various user needs. These can include the time series containing BOLD components expressed in ARU and the time series containing BOLD components. The time series can be used as inputs for conventional task-based (e.g., general linear model) and / or resting-state (e.g., functional connectivity) analyses.

[0061] DICOM compliant spatial maps of the BOLD components from ME-ICA may be output individually and expressed in quantitative ARU. These maps reflect individual-specific networks of functional brain activity.

[0062] PDF compliant reports summarizing data quality and network activity may be output alongside images. A report may be generated for each ME-fMRI series and can contain the following information:• Operational Data o Exam identifier: A QR code pertaining to the subject and exam; o Exam parameters: date, time, scanner make and model, type of fMRI (i.e., taskbased or resting-state), and study / sponsor; o Data quality descriptors: metrics summarizing the amount of motion (i.e., framewise displacement) and signal sensitivity (i.e., temporal signal -to-noise ratio) with corresponding pass / fail status based on established criteria.• Neurobiological data o Network Name The field-established name for the functional network based on literature of associated tasks; o Spatial Maps of Functional Network Expression: Glass brain renderings (population-reference anatomy in standard space) of activity for each canonical network; o Summary value of network activity magnitude: The values of activity in ri / ?2 units from a map are summarized statistically using a function such as the mean or a chosen percentile (e.g, 95%) to represent the level of activity of the network to compare against reference ranges; o Ranges of healthy functional network expression: A reference range for healthy expression for each functional network in ri / ?2and where the individual falls on the range;o Qualitative rating of activity magnitude: A qualitative label (e.g., high, normal, low) of the summary value of the network’s activity relative to the healthy matched population activity value.

[0063] The images generated after processing and accompanying reports may be transferred through RIS to a PACS system. A clinician may access the images using standard PACS tools and visualization capabilities built into the PACS system. Visualization, manipulation, and annotation will all be part of the installed PACS system. All functional images will have the capability to be overlaid atop anatomical images in standard DICOM viewers.

[0064] Figure 2 is a block diagram of an exemplary apparatus 200 suitable for implementing the method of Figure 1. The apparatus 200 may take the form of an application server provided by the same facility operating the MRI apparatus.

[0065] The apparatus 200 includes a processor 204. The apparatus 200 also includes a storage 208 coupled to the processor 204. The storage 208 comprises a set of program instructions in the form of a plurality of subsystems, configured to be executed by the processor 204.

[0066] The processor(s) 204, as used herein, means any type of computational circuit, such as, but not limited to, a microprocessor, a microcontroller, a complex instruction set computing microprocessor, a reduced instruction set computing microprocessor, a very long instruction word microprocessor, an explicitly parallel instruction computing microprocessor, a digital signal processor, or any other type of processing circuit, or a combination thereof. In one embodiment, processor 204 is an Intel Xeon or AMD Epyc server processor with at least 8 cores, clocked at Epic least 1.5 GHz, and having an L2 cache exceeding 1 MB.

[0067] The storage 208 includes a plurality of subsystems stored in the form of executable programs which instructs the processor 204 to perform the method steps. The plurality of subsystems includes: a preprocessor subsystem 212, a PCA module subsystem 216, an ICA module subsystem 220, an ARU conversion subsystem 224, and an output generator subsystem 228.

[0068] Computer memory elements implementing the storage 208 may include any suitable memory device(s) for storing data and executable program, such as read only memory, random access memory, erasable programmable read only memory, electrically erasable programmable read only memory, hard drive, removable media drive for handling memory cards and the like. Embodiments of the present subject matter may be implemented in conjunction with program modules, including functions, procedures, data structures, and application programs, forperforming tasks, or defining abstract data types or low-level hardware contexts. Executable programs stored on any of the above-mentioned storage media may be executable by the processor(s) 204.

[0069] The program instructions may be deployed via Docker, which has several advantages including a wide basis of support in the healthcare ecosystem (and more broadly), traceability with precise version control and provenance, security, and ease of software installation.

[0070] Although Figure 2 illustrates the apparatus 200 including a user interface 236, one skilled in the art can envision that the apparatus 200 can be connected to several user devices (not shown) located at different locations via the network interface 232.

[0071] The plurality of subsystems includes a preprocessor subsystem 212. The preprocessor subsystem 212 is capable of performing a variety of tasks with BOLD time series data. The preprocessor subsystem 212 may combine BOLD time series data taken at a plurality of TEs to produce a combined image time series prior to further processing. Preprocessing may also include basic steps to keep the timeseries from all TEs in alignment prior to combination and denoising. These steps include rigid body motion correction, T2* -driven functional-anatomical co-regi strati on, de-obliquing, and slice-time correction. The combination of BOLD time series data may involve a weighted sum of signals across TEs, as discussed above.

[0072] The denoiser module subsystem 216 is configured to perform principal components analysis on combined image time series data to remove noise signals that typically have low variance and are Gaussian distributed.

[0073] The plurality of subsystems also includes a neural signal identifier subsystem 220. The neural signal identifier subsystem 220 is configured to process the denoised data using ICA and use heuristics to differentiate between BOLD signals and non-BOLD signals. The time series data is reassembled with only BOLD signals for further processing and spatial maps may be generated as well.

[0074] The plurality of subsystems also includes an ARU conversion subsystem 224. The ARU categorizer subsystem 224 is configured to take reassembled time series data from the neural signal identifier subsystem 220 and convert it to autocalibrated relaxometry units (ARUs) as discussed above.

[0075] The report generator module 228 may be used to generate a variety of reports summarizing data quality and network activity as discussed above.

[0076] The network interface 232 provides connectivity to other RIS assets on a computer network, such as PACS and an MRI scanner. The network interface 232 can be used to receive MRI data from RIS assets and to transmit images, reports and data to a PACS system.

[0077] The user interface 236 can receive input from a user to control the operation of the system and display output related to that operation as well as images, reports, and data.

[0078] Figure 3 shows the embodiment of Figure 2 deployed as a network appliance 200 operating on a network that is connected to RIS assets such as an MRI scanner 300, a data store 304 containing fMRI imagery, and PACS endpoint 312. From endpoints on such a network, such as RIS terminal or PACS viewers, images can be sent to the appliance 200, which will trigger analysis, after the completion of which the results will be sent to a configured PACS endpoint 312.

[0079] Network appliance 200 is a DICOM compliant system with functionality to interact with PACS systems including receiving and sending images using C-STORE calls to and from the system, respectively. Inputs and outputs of the system are in DICOM format to operate on clinical radiologic imaging networks. The network appliance 200 also presents an interface to DICOM networks with which the DICOM images are pushed to it using standard Radiology Information System (RIS) procedures.

[0080] The network appliance 200 does not require continuous internet connectivity to conduct its clinical functionality. However, the network appliance 200 can utilize internet connectivity to facilitate, e.g., technical support, and the appliance 200 can be configured to access a preconfigured remote Internet endpoint, for which IP address, authentication, and other connectivity information may be preset, allowing local hospital firewall rules to allow routes out.Equivalents

[0081] The methods, systems, and devices discussed above are examples. Various configurations may omit, substitute, or add various procedures or components as appropriate. For instance, in alternative configurations, the methods may be performed in an order different from that described, and that various steps may be added, omitted, or combined. Also, features described with respect to certain configurations may be combined in various other configurations. Different aspects and elements of the configurations may be combined in a similar manner. Also, technology evolves and, thus, many of the elements are examples and do not limit the scope of the disclosure or claims.

[0082] Embodiments of the present disclosure, for example, are described above with reference to block diagrams and / or operational illustrations of methods, systems, and computerprogram products according to embodiments of the present disclosure. The functions / acts noted in the blocks may occur out of the order as shown in any flowchart. For example, two blocks shown in succession may in fact be executed substantially concurrent or the blocks may sometimes be executed in the reverse order, depending upon the functionality / acts involved. Additionally, or alternatively, not all of the blocks shown in any flowchart need to be performed and / or executed. For example, if a given flowchart has five blocks containing functions / acts, it may be the case that only three of the five blocks are performed and / or executed. In this example, any of the three of the five blocks may be performed and / or executed.

[0083] A statement that a value exceeds (or is more than) a first threshold value is equivalent to a statement that the value meets or exceeds a second threshold value that is slightly greater than the first threshold value, e.g., the second threshold value being one value higher than the first threshold value in the resolution of a relevant system. A statement that a value is less than (or is within) a first threshold value is equivalent to a statement that the value is less than or equal to a second threshold value that is slightly lower than the first threshold value, e.g., the second threshold value being one value lower than the first threshold value in the resolution of the relevant system.

[0084] Specific details are given in the description to provide a thorough understanding of example configurations (including implementations). However, configurations may be practiced without these specific details. This description provides example configurations only, and does not limit the scope, applicability, or configurations of the claims. Rather, the preceding description of the configurations will provide those skilled in the art with an enabling description for implementing described techniques. Various changes may be made in the function and arrangement of elements without departing from the spirit or scope of the disclosure.

[0085] Having described several example configurations, various modifications, alternative constructions, and equivalents may be used without departing from the spirit of the disclosure. For example, the above elements may be components of a larger system, wherein other rules may take precedence over or otherwise modify the application of various implementations or techniques of the present disclosure. Also, a number of steps may be undertaken before, during, or after the above elements are considered.

Claims

CLAIMSWhat is claimed is:

1. A method for auto-calibrated functional magnetic resonance imaging, the method comprising: receiving volumetric time series data describing the blood-oxygen-level-dependent (BOLD) values of each of a plurality of voxels in a brain at a plurality of echo times; computing a combined time series data using a weighted sum of the values of each of the voxels across the plurality of echo times; decomposing the combined time series data using independent component analysis to identify neural signals; converting the identified neural signals into units of / ?2,and providing an output incorporating the identified neural signals.

2. The method of claim 1 wherein the volumetric time series data is received from a magnetic resonance imaging scanner or a data store.

3. The method of claim 1 further comprising preprocessing the volumetric time series data.

4. The method of claim 3 wherein preprocessing comprises at least one of include rigid body motion correction, T2-driven functional-anatomical co-registration, de-obliquing, or slicetime correction.

5. The method of claim 1 wherein the coefficients of the weighted sum are where TEtis an echo time of interest, n is the number of theplurality of echo times, and T2is the susceptibility -weighted transverse relaxation contrast.

6. The method of claim 1 further comprising applying principal components analysis to the combined time series data to remove noise.

7. The method of claim 1 further comprising identifying neural signals using ICA and TE- dependence metrics to identify components with significant BOLD weighting and nonsignificant non-BOLD weighting.

8. The method of claim 1 wherein providing an output comprises generating reassembled volumetric time series data from components with significant BOLD weighting and nonsignificant non-BOLD weighting while discarding components with non-significant BOLD weighting and with or without high non-BOLD weighting.

9. The method of claim 1 wherein the output is at least one of an anatomical MRI image, a spatial map of an identified neural signal, a plurality of denoised time series images, or a report summarizing data quality and network activity.

10. The method of claim 1 wherein the volumetric time series data comprises at least one of: data from a single subject taken at different times; data from a plurality of subjects taken at the same time or the same site; unprocessed data from a plurality of scanners, wherein at least two of the plurality are different models of scanner.

11. A system for auto-calibrated functional magnetic resonance imaging, the system comprising: a processor; a user interface; and a memory storing instructions that, when executed by the processor, cause the processor to: receive volumetric time series data describing the blood-oxygen-level-dependent (BOLD) values of each of a plurality of voxels in a brain at a plurality of echo times; compute a combined time series data using a weighted sum of the values of each of the voxels across the plurality of echo times; decompose the combined time series data using independent component analysis to identify neural signals; convert the identified neural signals into units of R2, and provide an output incorporating the identified neural signals.

12. The system of claim 11 wherein the system further includes a network interface and the volumetric time series data is received from a magnetic resonance imaging scanner or a data store via the network interface.

13. The system of claim 11 wherein the memory further comprises instructions that, when executed by the processor, cause the processor to preprocess the volumetric time series data.

14. The system of claim 13 wherein preprocessing comprises at least one of include rigid body motion correction, T2* -driven functional-anatomical co-registration, de-obliquing, or slicetime correction.

15. The system of claim 11 wherein the coefficients of the weighted sum are where TEtis an echo time of interest, n is the number of theplurality of echo times, and T2* is the susceptibility-weighted transverse relaxation contrast.

16. The system of claim 11 wherein the memory further comprises instructions that, when executed by the processor, cause the processor to apply principal components analysis to the combined time series data to remove noise.

17. The system of claim 11 wherein the memory further comprises instructions that, when executed by the processor, cause the processor to identify neural signals using ICA and TE- dependence metrics to identify components with significant BOLD weighting and nonsignificant non-BOLD weighting.

18. The system of claim 11 wherein providing an output comprises generating reassembled volumetric time series data from components with significant BOLD weighting and nonsignificant non-BOLD weighting while discarding components with non-significant BOLD weighting and with or without high non-BOLD weighting.

19. The system of claim 11 wherein the output is at least one of an anatomical MRI image, a spatial map of an identified neural signal, a plurality of denoised time series images, or a report summarizing data quality and network activity.

20. The system of claim 11 wherein the volumetric time series data comprises at least one of: data from a single subject taken at different times; data from a plurality of subjects taken at the same time or the same site; unprocessed data from a plurality of scanners, wherein at least two of the plurality are different models of scanner.

Images