How to harmonize data between machines.

Abstract

A method of harmonizing data from a new machine of a first type with data from a reference machine of a second type, the method including: receiving a new frequency domain spectrum of a process performed by the new machine; and determining a harmonized new frequency domain spectrum of the process performed by the new machine by applying a set of harmonized new machine specific weights to the new frequency domain spectrum, the set of harmonized new machine specific weights being determined based on a filter frequency domain spectrum of a hardware filter of the new machine and a filter frequency domain spectrum of a hardware filter of the reference machine. The harmonized new frequency domain spectrum is harmonized to a corresponding frequency domain spectrum of the process when the process is performed by the reference machine.

Description

[Technical field]

[0001] FIELD OF THEINVENTION The present invention relates to a method and system for reconciling data between different machines, and in particular, but not exclusively, to a method and system for reconciling data generated by different machines having different hardware filters, and has particular application to EEG data. [Background technology]

[0002] background Electroencephalography (EEG) is a widely used imaging tool in neuroscience, especially due to its low cost and portability. EEG is an electrophysiological monitoring technique that records electrical activity on the scalp to determine the macroscopic activity of the brain's surface layers below. Typically, electrodes are placed on the scalp to measure voltage fluctuations resulting from the brain's electrical activity. EEG can be used to diagnose epilepsy, dementia, sleep disorders, tumors, seizures, depth of anesthesia and coma, and many other conditions.

[0003] Many different EEG devices are available to cover a wide range of uses for research and routine assessment. However, comparing EEG data collected from different devices is not straightforward. Methods optimized for one machine type may not be transferable to other machine types, as the differences in results may be much larger than the effects under investigation, thus making it difficult to validate results on different data sets. The main differences between different machine types arise from the hardware filters used during data recording. This leads to differences between data collected from different machine types (especially in the very low and very high frequency ranges). This problem is also found in other types of machines and is not limited to EEG machines.

[0004] It is known that for time domain analysis, software filters can be used to attempt comparability across various machine types. The usefulness of these filters is limited to some time domain analysis and visual inspection of the data. These software filters are inadequate for frequency domain analysis involving frequency ranges affected by the machine's hardware filters. This is because these software filters change the power spectrum in such a way that the analysis may no longer be meaningful. The present invention has been devised in light of the above considerations. Summary of the Invention [Means for solving the problem]

[0005] Summary of the Invention Generally, some aspects of the present invention provide methods and systems for harmonizing the spectral content of data across multiple machine types by converting data from a new machine of a first type into data of a reference machine of a second type by applying machine type specific weightings to the spectral content data of the new machine.

[0006] According to a first aspect of the present invention, there is provided a method of reconciling data from a new machine of a first type with data from a reference machine of a second type, the method comprising: receiving a new frequency domain spectrum of a process performed by the new machine; and determining a harmonised new frequency domain spectrum of the process performed by the new machine, where the harmonised new frequency domain spectrum is harmonized with a corresponding frequency domain spectrum of the process when performed by the reference machine by applying a set of harmonising new-machine specific weights to the new frequency domain spectrum, where the set of harmonising new-machine specific weights was determined based on the filter frequency domain spectrum of the hardware filter of the new machine and the filter frequency domain spectrum of the hardware filter of the reference machine.

[0007] In this way, the harmonized new frequency domain spectrum of the process performed by the new machine is harmonized with the frequency domain spectrum of the same process performed by the reference machine (particularly by possessing the filter characteristics of the hardware filter of the reference machine). Thus, data from different machines can be pooled to allow comparability of data sets recorded by different machines. Thus, analytical results can be verified on different data sets, and data from the same experiment or process can be collected on different machines for analysis. This can increase the number of data sets used in a study, thus improving the accuracy and efficiency of such studies, and improving the logistics of large-scale studies. Furthermore, this harmonization of the spectral content of the data across different machine types is suitable for network analysis.

[0008] Simulations described below in the detailed description demonstrate that methods according to embodiments of this and following aspects precisely enable matching of frequency domain data from a new machine with data from a reference machine despite various hardware filter characteristics.

[0009] The following describes optional features, which may be applied alone or in any combination with any aspect of the present invention.

[0010] As used herein, a type of machine may refer to a type of hardware filter used in the same machine. Thus, a first type of machine and a second type of machine may have different hardware filters. For example, a first type of machine may each have a first type of hardware filter, and a second type of machine may each have a second type of hardware filter, where the second type of hardware filter is different from the first type of hardware filter. The first type of machine and the second type of machine may be manufactured by different entities and / or may collect and process data in different ways.

[0011] As used herein, a process may be any measurement process and / or data collection process, where data is measured / collected by a machine with hardware filters. A process may involve measuring and / or collecting data from a subject, such as a human subject. A process may involve measuring electrical activity in the brain (e.g., an electroencephalography (EEG) process). However, the above method of reconciling data from a new machine of a first type with data from a reference machine of a second type is not limited to EEG data; i.e., any data collected by machines with various hardware filters may be reconciled by the methods described herein.

[0012] The set of harmonized new machine specific weights may have been determined at least in part by dividing the filter frequency domain spectrum of the hardware filter of the reference machine by the filter frequency domain spectrum of the hardware filter of the new machine.

[0013] In some applications, the frequency domain spectrum of the hardware filters of the reference machine and / or the frequency domain spectrum of the hardware filters of the new machine may be known. In these applications, a set of harmonized new machine specific weights may be determined based on the known filter frequency domain spectrum of the reference machine and / or the known frequency domain spectrum of the new machine.

[0014] However, in many applications, the filter frequency domain spectrum of the reference machine and / or the new machine may not be known. In these situations, a set of harmonized new machine-specific weights may be determined based on the estimated filter frequency domain spectrum of the reference machine and / or the estimated filter frequency domain spectrum of the new machine. In particular, the filter frequency domain spectrum of the hardware filter of the new machine and / or the filter frequency domain spectrum of the hardware filter of the reference machine (on which the set of harmonized new machine-specific weights is based) may be the estimated filter frequency domain spectrum of the hardware filter of the new machine and / or the estimated filter frequency domain spectrum of the hardware filter of the reference machine, respectively.

[0015] In particular, the estimated filter frequency domain spectrum of the hardware filter of the reference machine may be based on a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type; and / or the estimated filter frequency domain spectrum of the hardware filter of the new machine may be based on a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a first type.

[0016] The one or more machines of the first type may include a new machine. Each of the frequency domain spectra in the set corresponding to the one or more machines of the first type may be for a different process.

[0017] The one or more machines of the second type may include a reference machine. Each of the frequency domain spectra in the set corresponding to the one or more machines of the second type may be for a different process.

[0018] Preferably, the set of frequency domain spectra of a plurality of different processes performed by one or more machines of a first type comprises 30 or more frequency domain spectra, more preferably the set comprises 40 or more, 50 or more, 100 or more, 200 or more, etc. frequency domain spectra.

[0019] Similarly, the set of frequency domain spectra of the different processes performed by one or more machines of the second type may include 30 or more frequency domain spectra, more preferably the set includes 40 or more, 50 or more, 100 or more, 200 or more, etc. frequency domain spectra.

[0020] The inventors have found that a set of at least 30 frequency domain spectra provides a stable approximation / estimation of each filter frequency domain spectrum. Increasing the number of frequency domain spectra beyond 30 further increases the reliability.

[0021] The filter frequency domain spectrum of the hardware filter of the new machine may be based on the average (e.g., median) of multiple frequency domain spectra of multiple different processes performed by one or more machines of the first type. In other words, the filter frequency domain spectrum of the hardware filter of the new machine may have been determined by taking the average (e.g., median) of multiple frequency domain spectra of multiple different processes performed by one or more machines of the first type.

[0022] Similarly, the filter frequency domain spectrum of the hardware filter of the reference machine may be based on the average (e.g., median) of multiple frequency domain spectra of multiple different processes performed by one or more machines of the second type. In other words, the filter frequency domain spectrum of the hardware filter of the reference machine may have been determined by taking the average (e.g., median) of multiple frequency domain spectra of multiple different processes performed by one or more machines of the second type.

[0023] In this way, the approximations / estimates of the filter frequency domain spectrum of the hardware filter of the reference machine and the new machine can be used as the filter frequency domain spectrum of the hardware filter of the reference machine and the filter frequency domain spectrum of the hardware filter of the new machine, respectively. The inventors have discovered that with harmonising weights applied to the new frequency domain spectrum, common spectral characteristics of the different processes cancel out when the estimated filter frequency domain spectrum of the reference machine is divided by the estimated filter frequency domain spectrum of the new machine. Thus, although the approximated filter frequency domain spectrum may differ from the actual (e.g., true) filter frequency domain spectrum, the common spectral characteristics of the processes cancel out when the harmonising weights are obtained.

[0024] Optionally, the method may include interpolating the harmonized new machine specific weights across frequencies, in this way the harmonized new machine specific weights are available for the same frequency bins as the new frequency domain spectrum.

[0025] The new frequency domain spectrum may be received from an external network. In some examples, the new frequency domain spectrum may be received (directly) from the new machine. Alternatively, the new frequency domain spectrum may be received from an external storage device or from local storage.

[0026] In some examples, the new frequency domain spectrum of the process performed by the new machine may be one of a set of frequency domain spectra of a number of different processes performed by one or more machines of the first type. In this way, the new frequency domain spectrum of the process performed by the new machine itself forms part of the set of frequency domain spectra used to calculate an estimated filter frequency domain spectrum of the hardware filter of the new machine.

[0027] Optionally, the method may include receiving time series data of the process performed by the new machine, and transforming the time series data into a new frequency domain spectrum of the process performed by the new machine (e.g., by Fourier filtering on the original Fourier transform).

[0028] The time series data may be received, for example, from the new machine. Alternatively, the time series data may be received from an external network, from an external storage device, or from local storage.

[0029] As used herein, a frequency domain spectrum refers to a representation of one or more signals in each particular frequency band across a range of frequencies. The term "frequency domain spectrum" may be used interchangeably herein with the term "power spectrum" or the term "periodogram," which is an estimator of the power spectrum. Thus, the new frequency domain spectrum of the process performed by the new machine may be a new periodogram. The new periodogram may be determined by transforming the time series data from the new machine into the frequency domain (e.g., by Fourier filtering) and then taking the absolute square value of the resulting raw frequency domain spectrum.

[0030] Where the method includes transforming the time series data into a new frequency domain spectrum of the process performed by the new machine in the original Fourier transform, the method may include recovering harmonized time domain data from the harmonized new frequency domain spectrum. For example, recovering the harmonized time domain data may include taking the inverse Fourier transform of the square root of the harmonized new frequency domain spectrum (along with the phase of the original Fourier transform of the time series data). Thus, harmonized time domain data of the process performed by the new machine (which is harmonized to and therefore equivalent to the time domain data of the process when performed by the reference machine) may be obtained.

[0031] The harmonized time domain data from the harmonized new frequency domain spectrum (i.e., the harmonized data from the new machine) can then be pooled with the corresponding time domain data from the reference machine, since this data has been harmonized and is therefore equivalent to the reference machine data.

[0032] The data may be electroencephalographic (EEG) data. In particular, the reference machine and the new machine may be EEG machines (e.g., for taking EEG measurements). However, in other embodiments, the data may be other types of data (e.g., measured by another medical monitoring / recording process).

[0033] The above method can also be applied to another machine (i.e., a second, third, fourth, etc. new machine) In this way, data between multiple machines can be harmonized.

[0034] The method may be computer-implemented. For example, the method may be implemented on one or more computing devices. The one or more computing devices may be, for example, one or more computers, servers, cloud-based devices.

[0035] The method of this aspect may include any combination of some, all, or none of the preferred and optional features described above.

[0036] In a second aspect, there is provided a method of generating a set of harmonized new machine specific weights for harmonizing data from a new machine of a first type and data from a reference machine of a second type, the method comprising: The method includes determining a set of harmonized new machine specific weights based on a filter frequency domain spectrum of the hardware filter of the reference machine and a filter frequency domain spectrum of the hardware filter of the new machine.

[0037] Optionally, determining the set of harmonized new machine specific weights may include dividing a filter frequency domain spectrum of a hardware filter of the reference machine by a filter frequency domain spectrum of a hardware filter of the new machine.

[0038] The method further comprises: receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a first type; and The method may include determining a filter frequency domain spectrum of a hardware filter of the new machine based on a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the first type.

[0039] The method further comprises: receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type; and It may include determining a filter frequency domain spectrum of a hardware filter of a reference machine based on a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type.

[0040] The filter frequency domain spectrum of the hardware filter of the new machine may be determined, at least in part, by averaging (e.g., taking the median) a set of frequency domain spectra of multiple different processes performed by one or more machines of the first type.

[0041] The filter frequency domain spectrum of the hardware filter of the reference machine may be determined at least in part by averaging (e.g., taking the median) a set of frequency domain spectra of multiple different processes performed by one or more machines of a second type.

[0042] The one or more machines of the first type may include a new machine. Each of the frequency domain spectra in the set corresponding to the one or more machines of the first type may be for a different process.

[0043] The one or more machines of the second type may include a reference machine. Each of the frequency domain spectra in the set corresponding to the one or more machines of the second type may be for a different process.

[0044] Preferably, the set of frequency domain spectra of a plurality of different processes performed by one or more machines of a first type comprises 30 or more frequency domain spectra, more preferably the set comprises 40 or more, 50 or more, 100 or more, 200 or more, etc. frequency domain spectra.

[0045] Similarly, the set of frequency domain spectra of the different processes performed by one or more machines of the second type includes 30 or more frequency domain spectra, more preferably the set includes 40 or more, 50 or more, 100 or more, 200 or more, etc. frequency domain spectra.

[0046] The inventors have found that a set of at least 30 frequency domain spectra provides a stable approximation / estimation of each filter frequency domain spectrum. Increasing the number of frequency domain spectra beyond 30 further increases the reliability.

[0047] The sets of frequency domain spectrum may be received from an external network, in some examples, the sets of frequency domain spectrum may be received from an external storage device, from local storage, or directly from one or more machines of the first / second type.

[0048] The method of the second aspect may be a computer-implemented method.

[0049] The new machine-specific weightings generated in the method of the second aspect may be used as new machine-specific weightings in the method of the first aspect.

[0050] Thus, according to a third aspect, there is provided a method of reconciling data from a new machine of a first type and data from a reference machine of a second type, the method comprising: receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a first type; receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type; determining a filter frequency domain spectrum of the hardware filter of the new machine by averaging a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the first type; determining a filter frequency domain spectrum of the hardware filter of the reference machine by averaging a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type; determining a set of harmonized new machine specific weights based on a filter frequency domain spectrum of the hardware filter of the reference machine and a filter frequency domain spectrum of the hardware filter of the new machine; receiving a new frequency domain spectrum of a process performed by the new machine; and The method includes determining a harmonized new frequency domain spectrum of the process performed by the new machine, where the harmonized new frequency domain spectrum is harmonized with a corresponding frequency domain spectrum of the process when performed by the reference machine by applying a set of harmonized new machine specific weights to the new frequency domain spectrum.

[0051] The method of the third aspect may be a computer-implemented method.

[0052] In a fourth aspect, the present invention provides a device for harmonizing data from a new machine of a first type and data from a reference machine of a second type, the device including a processor and a memory, the memory including machine executable instructions which, when executed on the processor, cause the processor to receive a new frequency domain spectrum of a process performed by the new machine; and determine a harmonized new frequency domain spectrum of the process performed by the new machine: the harmonized new frequency domain spectrum is harmonized with a corresponding frequency domain spectrum of the process when the process is performed by the reference machine by applying a set of harmonized new machine-specific weightings to the new frequency domain spectrum, the set of harmonized new machine-specific weightings being determined based on a filter frequency domain spectrum of a hardware filter of the new machine and a filter frequency domain spectrum of a hardware filter of the reference machine.

[0053] The memory may include machine executable instructions that, when executed on a processor, cause the processor to perform a method of the first aspect including any one of the optional features described with reference thereto, or a combination thereof, so long as compatible.

[0054] In some embodiments, the device may be a computing device, such as, for example, one or more computers, servers, or cloud-based devices. The device may be configured to communicate with the reference device and / or the new device and / or one or more storage devices via wired and / or wireless connections. In some examples, the device may form part of the new machine and / or the reference machine.

[0055] The new frequency domain spectrum (and / or time series data corresponding to the frequency domain spectrum) of the process performed by the new machine may be obtained from an external network, may be obtained directly from the new machine, or may be obtained from local storage. Similarly, the set of frequency domain spectra of the different processes performed by one or more machines of a first type and the set of frequency domain spectra of the different processes performed by one or more machines of a second type (which may be used in determining the filter frequency domain spectra of the hardware filters of the reference machine and the new machine, respectively) may be obtained from an external network, may be obtained directly from one or more machines of the first and second types, respectively, and / or may be obtained from local storage.

[0056] Optionally, the device may include one or more input / output adapters for receiving data and / or sending harmonized data (e.g., a harmonized new frequency domain spectrum of a process performed by a new machine) back into the network.

[0057] In a fifth aspect, there is provided a device for generating a set of harmonized new machine specific weights for harmonizing data from a new machine of a first type and data from a reference machine of a second type, the device comprising a processor and a memory, the memory comprising machine executable instructions that, when executed on the processor, cause the processor to: receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a first type; receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type; determining a filter frequency domain spectrum of the hardware filter of the new machine by averaging a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the first type; determining a filter frequency domain spectrum of the hardware filter of the reference machine by averaging a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type; and A set of harmonized new machine specific weights is determined based on the filter frequency domain spectrum of the hardware filter of the reference machine and the filter frequency domain spectrum of the hardware filter of the new machine.

[0058] The new machine-specific weightings generated by the device of the fifth aspect may be used by the device of the fourth aspect as new machine-specific weightings.

[0059] The memory may include machine-executable instructions that, when executed on a processor, may cause the processor to perform a method of the second aspect including any of the optional features described with reference thereto (or combinations thereof, so long as compatible).

[0060] In a sixth aspect, there is provided a system for reconciling data from a new machine of a first type with data from a reference machine of a second type, the system comprising one or more processors and a memory, the memory comprising machine executable instructions which, when executed on the processor, cause the processor to perform the method of the third aspect.

[0061] In a seventh aspect, there is provided a non-transitory computer-readable storage medium comprising machine-executable instructions which, when executed on a processor, cause the processor to perform a method of the first aspect, the second aspect and / or the third aspect (including any (or any combination so long as compatible) of the optional features described with reference thereto).

[0062] Further aspects of the invention include: a computer program comprising code which, when executed on a computer, causes the computer to perform the method of the first aspect, the second aspect and / or the third aspect; a computer readable storage medium storing a computer program comprising code which, when executed on a computer, causes the computer to perform the method of the first aspect, the second aspect and / or the third aspect; and a computer system programmed to perform the method of the first aspect, the second aspect and / or the third aspect.

[0063] The present invention includes combinations of the described aspects and preferred features unless such combinations of aspects are clearly impermissible or clearly avoided.

[0064] Diagram Overview Next, some embodiments and experiments illustrating the principles of the present invention will be discussed with reference to the accompanying drawings. [Brief description of the drawings]

[0065] [Figure 1A] 1 is a plot showing distributions of example network measures for healthy and diseased groups collected on the same EEG machine. [Figure 1B] 13 is a plot showing distributions of exemplary network measures for healthy and diseased groups collected on the same EEG machine and distributions of exemplary network measures for the healthy group collected on different EEG machines of different types. [Diagram 2] FIG. 13 is a diagram showing the results of rPDC analysis on original simulation data, filtered data, and data obtained by reversing the filter effect. [Diagram 3] 13 is a graph showing a comparison of uniformly distributed peak frequencies with a known filter power spectrum and a reconstructed filter power spectrum in a simulation from 100 data sets. [Figure 4] 1 is a flow diagram of a method for reconciling data from a new machine with data from a reference machine. [Diagram 5]FIG. 5 is a diagram illustrating an implementation of the method shown in FIG. 4. [Figure 6] Included are two graphs showing the results of a simulation to determine an approximated filter power spectrum compared to a true filter power spectrum. [Figure 7] 1 is a graph showing a comparison of a harmonization curve (a set of weightings) obtained from a set of estimated power spectra and a true harmonization curve derived from a known filter power spectrum. [Figure 8A] 13 is a graph of approximate filter values ​​(eg, median periodogram) at 10 Hz with error bars for increasing numbers of data obtained from a simulation process. [Figure 8B] 13 is a graph of approximate filter values ​​(eg, median periodograms) at 10 Hz with error bars for increasing numbers of data sets obtained from EEG data. [Figure 9] 1 is a graph of approximate filter values ​​(eg, median periodogram) at 10 Hz with error bars for 50 iterations of randomly selecting 30 out of 190 power spectra obtained from EEG data. [Figure 10] Included are two plots of the power spectra of the filters of the new and reference machines obtained from the simulation process. [Figure 11] 1 is a graph of a theoretical harmonization curve obtained in a simulation process. [Figure 12] Included are two plots of the approximate power spectra of the filters of the new and reference machines obtained in the data harmonization method. [Figure 13] 13 is a plot comparing theoretical weightings corresponding to the harmonization curve of FIG. 11 with weightings derived from the approximated power spectrum of FIG. 12. [Figure 14] FIG. 13 is a diagram showing the results of an rPDC analysis on original simulation data from a new machine, a reference machine, and a new machine harmonized with the reference machine. [Figure 15]1 is a graph of rPDC values ​​obtained from simulations of the new machine type, the reference machine type and the harmonized reference machine type. [Figure 16] Included are two plots of rPDC data for comparison of results with the data harmonization process versus results without the data harmonization process such as in FIG. [Figure 17] A plot showing a distribution of exemplary network measures for a healthy group and a diseased group collected on the same EEG machine, as well as exemplary network measures for the healthy group collected on different machines of different machine types, where the distribution for the healthy group has been harmonized according to a data harmonization process such as that of FIG. [Figure 18] FIG. 13 is a diagram of the grand average ERP from the new machine used to collect EEG data for the ERP experiments. [Figure 19] FIG. 13 is a plot of the grand average ERP from the new machine used to collect EEG data (data harmonized with the reference machine) for the ERP experiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0066] Detailed Description of the Invention Aspects and embodiments of the present invention will now be discussed with reference to the accompanying drawings. Further aspects and embodiments will be apparent to those skilled in the art. All documents mentioned in this text are incorporated herein by reference.

[0067] 1A is a plot showing how example EEG network measures for a single machine type (machine type 1) differ at the group level for a healthy group 10 compared to a sick group 12. In this illustrative example, the network measures used are frontal-out-degree, which is based on a directed measure, and frontal-out-degree, which measures the average connectivity strength away from the frontal electrodes.

[0068] However, when the same analysis is applied to EEG recordings with a different machine type (machine type 2), the network measures are different. This is shown in FIG. 1B, which shows the same fontal-out-degree network measures (same as those measured with machine type 1) for healthy group 10 and sick group 12, but now also shows the distribution of network measures for healthy group 14 recorded with machine type 2. As shown in FIG. 1B, the distributions of network measures for the healthy group measured with machine type 1 and machine type 2 are different (i.e. plot 10 is different from plot 14). Notably, in this example, the difference between the two EEG machines is larger than the disease effect. This difference highlights why it is not possible to directly compare EEG measures derived from different EEG hardware.

[0069] Data such as EEG data from both the new machine (e.g., machine type 2) and the reference machine (e.g., machine type 1) will be subjected to machine-type specific hardware filters (each with its own characteristics). Thus, the power spectrum of the hardware filtering process (the "joint power spectrum") contains the power spectrum of the process and the power spectrum of the hardware filters (and in particular, the product of the power spectrum of the process and the power spectrum of the hardware filters). It is the different power spectra of the hardware filters that cause differences between the distributions of network measures for the same group when the data are from different machine types.

[0070] In order to harmonize data across different machine types in the frequency domain, the inventors have discovered that harmonization weights can be applied to the power spectrum of the hardware filtering process. The harmonization weights are obtained by dividing the power spectrum of the hardware filter of the reference machine by the power spectrum of the hardware filter of the new machine. The power spectrum of the hardware filtering process performed by the new machine is then transformed by multiplying it with the obtained harmonization weights. This harmonized power spectrum is then harmonized to possess the filter characteristics of the hardware filter of the reference machine. After the harmonization weights are applied, the harmonized data in the time domain can be retrieved from the harmonized power spectrum as long as the original data is received in the time domain such that the phase of the original Fourier transform (to the frequency domain) of the unharmonized data is available as described below.

[0071] The inventors performed simulations to demonstrate that it is possible to reverse the effect of a hardware filter on simulated data. This is first demonstrated by using a single machine type and simulated data constrained to a known filter power spectrum. In particular, if the filter power spectrum is known, its effect on the data can be reversed by dividing the integrated power spectrum (e.g., the power spectrum of the underlying process multiplied by the power spectrum of the hardware filter) by the filter power spectrum.

[0072] This process of reversing the filter effect preserves frequency domain measures such as renormalised partial directed coherence (rPDC), which has been demonstrated by using simulation data to show how hardware filters affect the rPDC values ​​of a connection, as described below.

[0073] A three-dimensional vector autoregressive process (VAR) with a model order of 2 (VAR[2]) was simulated and filters with known characteristics were applied to each of the three individual processes.

[0074] The vector autoregressive process is given by Equation 1:

number

number

[0075] The 3D VAR process was simulated with N = 20,000 data points per channel, which were chosen to represent 100 seconds of data collected with a sampling rate of 200 Hz.

[0076] The interaction structure of this system involves three processes, where process x1 influences process x2, which influences process x3. The influence of process x1 on process x3 is indirect, since it is mediated by process x2.

[0077] Next, a filter was simulated by applying an AR[1] process in the time domain to each time series of the simulation data.

[0078] This AR[1] process is defined in Equation 2: Y(t)=bY(t-1)+x(t) (2) where x(t) is the

number

[0079] The analytical solution for the power spectrum of this known filter is defined by Equation 3:

number

[0080] The value of b in Equation 3 defines the shape of the filter characteristic and is set to 0.9 in the present simulations and is defined in Equation 2 above.

[0081] The first step in the simulation process was to estimate the power spectrum by computing a periodogram for each dimension of the simulation data. To do this, the time series data was Fourier transformed to provide the raw data Fourier transform, the positive frequency values ​​were extracted, and the absolute square of the raw data Fourier transform (which is the periodogram) was taken. A filter spectrum was then calculated for each frequency bin present in the periodogram by using Equation 3 above. Each frequency bin of the periodogram was then divided by the value of the filter spectrum at the respective frequency, dimension by dimension. With these steps, the filter effect on the periodogram is reversed, and the result is the "corrected periodogram."

[0082] To obtain the time domain data corresponding to the corrected periodogram, the absolute value of the corrected periodogram and the phase of the raw data Fourier transform were used in an inverse Fourier transform.

[0083] In addition to the rPDC analysis performed on the original and filtered simulation data, an rPDC analysis was also performed on the simulation data obtained after reversing the filter effect. FIG. 2 shows the results of the rPDC analysis on the original simulation data, the filtered data, and the data obtained after reversing the filter effect. FIG. 2 shows that the curve of the data obtained after reversing the filter effect almost perfectly overlaps with the curve from the original simulation data, which indicates that the filter effect can be reversed by dividing the integrated power spectrum by the power spectrum of the filter. Thus, if the power spectrum of the hardware filter is known, it is possible to reverse the effect of the filter and thus the effect of pooling data from different machines with different hardware filters.

[0084] However, for most applications, the filter power spectrum is not known. If the process is a white noise process, the power spectrum of the filter can be inferred since the power spectrum of a white noise process is constant across frequencies. Similarly, if there are multiple power spectra of various processes filtered by the same filter and these processes have uniformly distributed spectral peaks, a good approximation of the filter spectrum can still be inferred by taking the average (e.g., median) of the estimated spectra. This is because the various spectral peaks may cancel each other out when the average (e.g., median) is taken, and what remains is the filter power spectrum that is common to all filtered processes.

[0085] We demonstrated this by simulating 100 univariate autoregressive processes of order 2 (AR[2]) with peak frequencies uniformly distributed across the entire frequency range. The AR[1] process described above was used to simulate a filter according to Equation 3. The median of the spectral estimates (e.g., smooth periodograms) at each frequency was then taken and compared to the known filter power spectrum. The results are shown in plot 20 of FIG. 3. FIG. 3 shows how there is good agreement between the median of the spectral estimates and the median of the known filter power spectrum, and confirms that the filter can be reconstructed from data if a set of many power spectra with uniformly distributed peak frequencies is available.

[0086] However, for many applications, the available data may not include white noise processes (or many power spectra from the same filter (e.g., from a single machine type) that contain uniformly distributed spectral peaks). Instead, data may be available from a variety of processes that share some spectral characteristics (e.g., all spectral characteristics have peaks within a certain frequency range). It is therefore more difficult to reconstruct the filter power spectrum by taking the median of the estimated spectra, since the various spectral peaks do not cancel each other out well when the median is taken. The result of taking the median is therefore a combination of the filter power spectrum and the common spectral characteristics of the processes. The inventors have discovered that increasing the number of power spectra in the data set can help alleviate this challenge, but only if the additional power spectra broaden the distribution of the spectral peaks. Nevertheless, in many applications, the goal is to harmonize data from the same or similar processes recorded by a variety of machines. Thus, the approximated filter power spectra from both the new and reference machines will contain the common spectral characteristics of the processes.

[0087] Figure 4 is a flow diagram of an implementation of a method for reconciling data from a new machine with data from a reference machine. Figure 5 is a diagram 30 that also illustrates this method. The methods illustrated in Figures 4 and 5 can be performed, for example, in a computing device.

[0088] In S101 of Figure 4, a set of frequency domain spectra of a number of different processes performed by one or more machines of a first type is received. The set of frequency domain spectra may be received, for example, from an external network or from local storage. In Figure 5, the set of power spectra is labeled "Reference Data Set" instead of "New Machine."

[0089] In S102 of Figure 4, a set of frequency domain spectra of a plurality of different processes performed by one or more machines of a second type is received. Again, this set of frequency domain spectra may be received, for example, from an external network or from local storage. Again, in Figure 5, the set of power spectra is labeled "Reference Data Set" instead of "Reference Machine."

[0090] S101 and S102 may be performed in any order or simultaneously.

[0091] In S103 of FIG. 4, the filter frequency domain spectrum of the hardware filter of the new machine is estimated by averaging (e.g., taking the median) a set of frequency domain spectra of multiple different processes performed by one or more machines of a first type.

[0092] Similarly, in S104 of FIG. 4, the filter frequency domain spectrum of the hardware filter of the reference machine is estimated by averaging (e.g., taking the median) a set of frequency domain spectra of multiple different processes performed by one or more machines of a second type.

[0093] For completeness, these determined filter power spectra (labeled "Approximate Filter Characteristics" in FIG. 5) for both the reference and new machines are approximations. In reality, they may differ from the true filter power spectra. However, this difference is taken into account when the weights are calculated later in the methods described herein.

[0094] S103 and S104 may be performed in any order or simultaneously.

[0095] In S105 of FIG. 4, a set of harmonized new machine-specific weights is determined based on the filter frequency domain spectrum of the hardware filter of the reference machine and the filter frequency domain spectrum of the hardware filter of the new machine. This set of harmonized new machine-specific weights is labeled as "harmonized curve" in FIG. 5. The set of weights is calculated by dividing the approximated filter power spectrum of the reference machine by the approximated filter power spectrum of the new machine. Common spectral characteristics of multiple different processes cancel out when the approximated reference hardware frequency domain spectrum is divided by the approximated new hardware frequency domain spectrum. Thus, although the approximated filter frequency domain spectrum may be very different from the actual (e.g., true) filter spectrum, the common spectral characteristics of the processes cancel out when the harmonized weights are calculated.

[0096] At S106, a new power spectrum of the process performed by the new machine is received. The new power spectrum of the process performed by the new machine can be received, for example, from an external network or from local storage.

[0097] In some examples, the new power spectrum of the process performed by the new machine may be obtained from time series data of the process performed by the new machine. In these examples, the method may include receiving the time series data of the process performed by the new machine and transforming the time series data into a new frequency domain spectrum of the process performed by the new machine in an original Fourier transform. The time series data may be received from an external network or from local storage. Once the new power spectrum of the process performed by the new machine is obtained, it may be stored locally so that it can be subsequently received at S106.

[0098] For completeness, in some examples, the new power spectrum of the process performed by the new machine may be one of the frequency domain spectra in the set of frequency domain spectra received in S101 (e.g., one of the frequency domain spectra of a plurality of different processes performed by one or more machines of a first type).

[0099] In S107, the weights are applied to the new power spectrum of the process performed by the new machine (labeled "Individual Subject Data Set of the new machine" in FIG. 5) to generate a harmonized new power spectrum (labeled "Harmonized Individual Subject Data Set" in FIG. 5). The new harmonized power spectrum of the new machine is equivalent to the power spectrum of the process when the same process is performed by the reference machine (because the effect of the hardware filters of the new machine has been harmonized with the effect of the hardware filters of the reference machine (see FIG. 5)). This data can therefore be pooled for further analysis. If necessary, the weights can be interpolated across frequencies so that they are available for the same frequency bins as the new power spectrum.

[0100] Optionally, if the new power spectrum of the process performed by the new machine is obtained from time series data, the method may also include transforming the harmonized new power spectrum into time domain data (by taking the inverse Fourier transform of the square root of the harmonized new power spectrum together with the phase of the original Fourier transform of the time series data), which may then be directly compared with the time domain data of the reference machine.

[0101] The method may only include S106 and S107 in Fig. 4 (e.g., if the set of harmonized new machine-specific weightings is already known). The method for generating the set of harmonized new machine-specific weightings is described in S101-S105 in Fig. 4.

[0102] As mentioned above, the determined filter power spectra of both the reference and new machines are only approximations. In reality, they may differ from the true filter power spectra of these filters. This is shown in FIG. 6, which shows two graphs 40 and 42 showing the simulation of 100 AR[2] processes with peak frequencies varying within a fixed frequency range. FIG. 6 shows the results of a simulation (i.e., recordings made for two different machine types) in which two different AR[1] processes (equation 3 above with b=0.9 and b=0.2 in graphs 40 and 42 of FIG. 6, respectively) to simulate two different filters were applied to each of the time series data of 50 time series. The resulting filter approximations are poor because they also contain frequency components common to all processes represented in the set of power spectra used to calculate the approximations.

[0103] However, when the filter harmonization weights are calculated, the common spectral characteristics cancel each other out. This is illustrated by the graph 50 shown in Figure 7, which shows a comparison of weights across frequency derived from the data with those calculated from the known true filter spectrum. This may also be defined as a "harmonization curve."

[0104] The inventors have found that it is desirable for the set of power spectra of the different processes performed by one or more machines of the same type as the reference machine and the set of power spectra of the different processes performed by one or more machines of the same type as the new machine to each contain at least 30 power spectra. In particular, the inventors have used an AR[2] process with a peak frequency in a specific frequency range to investigate how many power spectra are needed to obtain a stable approximation of the filter power spectrum. The median of the estimated spectrum (smoother periodogram with a smoothing width of 20 frequency bins) of the five power spectra is taken, and then the number of power spectra is increased to 200. FIG. 8A shows a graph 60 of the result median with error bars of one frequency. It can be observed that for very few power spectra in the set, the approximate filter value still fluctuates and thus has a large confidence interval indicating that it is not well defined. From a set of about 30 power spectra, the approximate filter value becomes stable, and increasing the number of data sets increases the confidence. Therefore, the present invention has concluded that a minimum of 30 power spectra in the data set is desirable.

[0105] This analysis was also repeated with EEG data measured by one EEG machine type to demonstrate that at least 30 power spectra in the data set are desirable. This analysis was based on data collected from 83 healthy subjects and 107 diseased subjects. The number of power spectra in the set for determining the filter power spectrum was randomly selected from a pool of 190 power spectra available. As in the simulation, the analysis began with 5 power spectra in the data set and was increased to the maximum number available of 190. FIG. 8B shows a plot 62 of the values ​​obtained for the approximated filter power spectrum at 10 Hz across the number of power spectra used in the data set for the approximation. Similar to the simulation results of FIG. 8A, the values ​​obtained for the approximated filter power spectrum out of the 30 power spectra in the data set become stable.

[0106] For completeness, we also explored whether this approximation would be similar if different power spectra were used for the approximation. We set the number of power spectra in the data set to 30 and randomly selected 50 times from the 190 power spectra available. The results of the approximated filter value at 10 Hz are shown in graph 64 of FIG. 9. Only small variations are seen, and it is therefore determined that 30 power spectra in the data set are sufficient to obtain a stable and reproducible approximation of the filter power spectrum. More power spectra in the data set would be desirable but not required.

[0107] The inventors have tested the above disclosed method of reconciling data from a new machine with data from a reference machine by simulation. The same three-dimensional system as described above (i.e. defined by Equation 1). Two different filters were applied to the same implementation of this system to generate data for the reference machine and data for the new machine. The power spectra of the two different filters (defined above in Equation 3) were applied to the same implementation of the three-dimensional system: where b=0.9 for the new machine and b=0.2 for the reference machine. Figure 10 shows plots 70 and 72 of the power spectra of the resulting filters for the new machine and the reference machine, respectively. Both spectra are plotted on a logarithmic scale.

[0108] When the power spectrum of the filter from the reference machine is divided by the power spectrum of the filter from the new machine, a theoretical weighting for harmonizing the data is obtained. This theoretical harmonization curve is shown in graph 74 of Figure 11. However, as discussed above, in most applications the power spectrum of the filter is unknown and therefore the harmonization curve needs to be estimated.

[0109] To demonstrate how to approximate the power spectrum of the filter, we simulated ten three-dimensional systems separately by using the same VAR coefficients described above in relation to Equation 1, and applied the simulated filters to the time series data of both the reference and new machines. This corresponds to simultaneously acquiring ten data sets from the same subject by both machines, and the ten data sets can then be used to approximate the filter power spectrum. The simulation of ten realizations of the three-dimensional system results in 30 estimated spectra (e.g., smoothed periodograms) for each machine of the two machines. A smoothing width of 20 frequency bins was used. The median of the 30 estimated spectra (ten realizations of each of the three original dimensions) was taken for the approximated power spectrum of the filter. Figure 12 shows plots 76 and 78 of the approximated power spectra of the new and reference machines, respectively. The approximated power spectra include a combination of the respective filter power spectra and the common spectral characteristics of the processes from which they were derived. Through division of the approximate power spectrum of the reference machine and the approximate power spectrum of the new machine, common spectral characteristics of the processes are cancelled out and a weighting is obtained which is applied to the integrated power spectrum of the new machine.

[0110] 13 shows a plot 80 of the resulting theoretical weights 82 and weights derived from the estimated power spectrum 84. As shown, the weights derived from the estimated power spectrum match closely to the theoretical weights.

[0111] The rPDC measure was also used to compare the connection strengths between processes from the reference machine, the new machine, and the new machine harmonized with the reference machine. FIG. 14 shows a diagram 86 of the results obtained from the rPDC analysis. For all three machine scenarios (e.g., the reference machine, the new machine, and the new machine harmonized with the reference machine), this analysis shows only non-zero rPDC values ​​of the connections present in the simulation. It can be seen that for the present connections (x1 to x2 and x2 to x3), the values ​​obtained from the new machine are very different from those of the reference machine. The reason for this difference is the different filters since the same original simulation data is used for this analysis. After applying the above harmonization procedure to the data from the new machine, the resulting values ​​are comparable to those obtained from the reference machine. FIG. 14 shows how the above method can be used to obtain comparable connection strengths since the rPDC values ​​of the reference machine and the new machine harmonized with the new machine match those of the reference machine.

[0112] An additional 500 realizations of the three-dimensional system (e.g., as defined by Equation 1) were simulated, and the reference filter was applied to each time series data. These 500 realizations were then simulated to generate confidence bands of rPDC values. The rPDC values ​​obtained from the harmonized new machine to the bands of rPDC values ​​of the reference machine. FIG. 15 shows a plot 88 of the rPDC values ​​obtained from the simulation of the connection from x2 to x3. The plot 88 in FIG. 15 shows that there is a large difference between the reference machine results (grey and dashed lines) and the new machine results (dotted lines). The harmonized new machine results (thick black lines) can also be compared to the reference machine results (grey and dashed lines). FIG. 15 shows that the harmonized new machine results overlap with the reference machine results, confirming that the above procedure is effective in obtaining equal connection strengths.

[0113] FIG. 16 shows two graphs; the first graph 90 shows the rPDC results of the new machine and the reference machine with its various filters; and the second graph 92 shows the rPDC results after harmonization. In the simulation, the connection strength of System A was chosen to be higher than System B. Graph 90 shows that the rPDC values ​​of the present system are higher at low frequencies than System B, while System A has higher rPDC values ​​at higher frequencies; thus indicating how the connection strength is incorrect at low frequencies where no harmonization is performed. In contrast, graph 92 shows that the rPDC values ​​of System A are consistently higher than System B, correctly indicating that the present system A has a stronger connection, and confirms that the harmonization procedure described above is effective in obtaining equal connection strengths even when the present systems have various hardware filters. Graph 92 further shows that even if it is not possible to reverse the filter effect, the relative connection strengths can be recovered by the harmonization procedure described above.

[0114] As mentioned above, the inventors also performed an analysis with EEG data. Figure 17 shows a corresponding plot 100 to that shown in Figure 1B, but with the above-mentioned harmonization method applied to data measured by machine type 2. Plot 100 shows that the distribution of network measures obtained from healthy group 114 measured by machine type 2 overlaps with the distribution of network measures obtained from healthy group 110 measured by machine type 1 (corresponding to curve 10 in the plot of Figure 1B). Also shown is sick group 112 measured by machine type 1 (corresponding to curve 12 in the plot of Figure 1B). In other words, the results from the two different machine types of healthy groups 110 and 114 become comparable after application of the above-mentioned harmonization method.

[0115] To verify that the above-mentioned method would also be effective when frequency domain results were transformed to the time domain, the inventors applied the above-mentioned method to data obtained from an experiment investigating event-related potentials (known time-domain analysis of EEG data). An event-related potential (ERP) is a measured brain response that is a direct consequence of a specific perceptual, cognitive or motor event. Data from an ERP experiment (showing differences in two different stimuli) were used to test that application of the proposed harmonization method described above would not impair conclusions drawn from the analysis of event-related potentials.

[0116] The ERP experiment investigated the standard old / new effect (HIT vs. correct rejection), where subjects were presented with a stimulus and asked to judge whether the stimulus was part of the study list (i.e., old) or not (i.e., new). Throughout the ERP experiment, EEG was recorded continuously by 64 electrodes on the scalp while subjects were performing the task. Subjects were shown objects that appeared in various locations on the screen during the encoding phase. Then, during the retrieval phase, subjects were shown the same objects and an additional random object that had not been shown before (i.e., new object). There were 12 object location encoding phases, each followed by a retrieval phase. Each encoding phase consisted of 12 trials. In addition to the 12 old objects shown during the encoding phase, 12 new objects were included in each retrieval phase. Thus, each retrieval phase consisted of 24 trials. This resulted in a total of 12 × 24 = 288 trials for each subject. During the retrieval phase, subjects were required to press a button to indicate where the object appeared during the encoding phase. The add button could be pressed by the subject if he recognized the object as one that was not presented during the encoding phase (i.e., a new object). During the retrieval phase, an attempt was considered a "HIT" if the subject correctly assigned the location where the object appeared during the encoding phase. An attempt was considered a "Correct Rejection" (CR) if the subject correctly recognized that the object was a new object that had not appeared in the encoding phase. EEG data for this ERP experiment were recorded from 59 healthy subjects.

[0117] The following standard ERP preprocessing was applied to the continuous EEG data recorded during the ERP experiment. The continuous data was first bandpass filtered between 0.1 Hz and 35 Hz. The data was then referenced to the average and binned into epochs based on several trials. Each trial epoch included the EEG time series data from the time the stimulus was presented to the subject (0 s) to 1 s after the stimulus was presented to the subject (1 s). Subjects with fewer than 16 trials in either the HIT or CR condition were excluded from the analysis to ensure that sufficient data was available for averaging. This resulted in 6 of the 59 subjects being excluded from the analysis. The amplitudes of all trials from all subjects were averaged to generate a grand average for each condition. The resulting ERP grand averages shown in Figure 18 show clear differences between the HIT and CR conditions from 0.4 s to 0.7 s.

[0118] To demonstrate that the harmonization method described above does not disrupt the ERPs, the above procedure was repeated after applying the harmonization method to the data. 190 data sets recorded by the reference machine were used to obtain the approximated filter power spectrum of the reference machine. The filter power spectrum of the new machine was approximated based on data from the 59 subjects who participated in the ERP experiment. However, data from the resting state segment preceding the ERP experiment was used to obtain the approximated filter power spectrum of the new machine. The harmonized data was then analyzed in the same way as the original data. The resulting ERP grand mean is shown in Figure 19.

[0119] It can be seen from Figures 18 and 19 that in both cases the two conditions (HIT and CR) are well separated. Furthermore, by comparing the Y-axis, after harmonization (as shown in Figure 19), the amplitude of the ERP grand average is lower. Different amplitudes may be caused by different EEG machine types with different amplitudes. This can be corrected by multiplying the grand average ERP by a fixed factor. The results in Figures 18 and 19 are similar, indicating that the harmonization method disclosed herein does not disturb the grand average ERP.

[0120] The systems and methods of the above embodiments, in addition to the structural components and user interactions described, may be implemented within a computer system (particularly within computer hardware or within computer software).

[0121] The term "computer system" includes hardware, software and data storage devices for implementing the system or performing the method according to the above-described embodiments. For example, a computer system may include a central processing unit (CPU), input means, output means and data storage. Preferably, the computer system has a monitor for providing a visual output display. The data storage may include RAM, disk drives or other computer readable media. A computer system may include multiple computing devices connected by a network and capable of communicating with each other over the network.

[0122] The methods of the above embodiments may be provided as a computer program arranged to perform the above-mentioned methods when executed on a computer, or as a computer program product or computer readable medium carrying the computer program.

[0123] The term "computer-readable medium" includes, without limitation, any non-transitory medium or media that can be read and accessed directly by a computer or computer system. Media may include, but are not limited to, magnetic storage media, such as floppy disks, hard disk storage media, and magnetic tape; optical storage media, such as optical disks or CD-ROMs; electrical storage media, such as memory, including RAM, ROM, and flash memory; and hybrids and combinations of the above, including magnetic / optical storage media.

[0124] The features disclosed in the foregoing description, or in the following claims, or in the accompanying drawings, and expressed in their specific form or in terms of means for performing a disclosed function, or methods or processes for obtaining a disclosed result, may be utilized, as appropriate, separately or in any combination of such features to realize the invention in various of its forms.

[0125] Although the present invention has been described in conjunction with the exemplary embodiments above, many equivalent modifications and variations will become apparent to those skilled in the art given this disclosure. Accordingly, the exemplary embodiments of the present invention described above are considered to be illustrative and not limiting. Various changes to the foregoing embodiments may be made without departing from the spirit and scope of the present invention.

[0126] In particular, although the methods of the above embodiments have been described as being implemented on the systems of the described embodiments, the methods and systems of the present invention need not be implemented in conjunction with each other, but may each be implemented on alternative systems or using alternative methods.

[0127] For the avoidance of any doubt, any theoretical explanations provided herein are provided for the purpose of enhancing the understanding of the reader, and the inventors do not wish to be bound by any of these theoretical explanations.

[0128] Any section headings used herein are for organizational purposes only and therefore should not be construed as limiting the subject matter described.

[0129] Throughout this specification, including the claims which follow, unless the context requires otherwise, the terms "comprise" and "include" or variations thereof such as conjugations thereof will be understood to imply the inclusion of a stated integer or step or group of integers or steps, but not the exclusion of any other integer or step or group of integers or steps.

[0130] It should be noted that, as used in this specification and the appended claims, the singular indefinite and definite articles include plural references unless the context clearly dictates otherwise. Ranges may be expressed herein as from "about" one particular value and / or to "about" another particular value. When such a range is expressed, another embodiment includes from the one particular value and / or to the other particular value. Similarly, when values ​​are expressed as approximations by use of the antecedent "about," it will be understood that the particular value forms another embodiment. The term "about" in connection with numerical values ​​is optional and means, for example, + / - 10%.

Claims

1. 1. A method of reconciling data from a new machine of a first type with data from a reference machine of a second type, the method comprising: receiving a new frequency domain spectrum of a process performed by the new machine; and determining a harmonized new frequency domain spectrum of the process performed by the new machine, wherein the harmonized new frequency domain spectrum is harmonized with a corresponding frequency domain spectrum of the process when performed by the reference machine by applying a set of harmonized new machine specific weights to the new frequency domain spectrum; The method wherein the set of harmonized new machine specific weights is determined based on a filter frequency domain spectrum of a hardware filter of the new machine and a filter frequency domain spectrum of a hardware filter of the reference machine.

2. 2. The method of claim 1, wherein the set of harmonized new machine-specific weights is determined at least in part by dividing the filter frequency domain spectrum of the hardware filter of the reference machine by the filter frequency domain spectrum of the hardware filter of the new machine.

3. the filter frequency domain spectrum of the hardware filter of the new machine is an estimated filter frequency domain spectrum of the hardware filter of the new machine; and / or The method of claim 1 or claim 2, wherein the filter frequency domain spectrum of the hardware filter of the reference machine is an estimated filter frequency domain spectrum of the hardware filter of the reference machine.

4. the estimated filter frequency domain spectrum of the hardware filter of the new machine is based on a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the first type; and / or 4. The method of claim 3, wherein the estimated filter frequency domain spectrum of the hardware filter of the reference machine is based on a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the second type.

5. the one or more machines of the first type include the new machine; and / or The method of claim 4 , wherein the one or more machines of the second type include the reference machine.

6. the set of frequency domain spectra of the plurality of different processes performed by the one or more machines of the first type includes at least 30 frequency domain spectra; and / or 6. The method of claim 5, wherein the set of frequency domain spectra of the plurality of different processes performed by the one or more machines of the second type comprises at least 30 frequency domain spectra.

7. the filter frequency domain spectrum of the new machine is based on an average of the set of frequency domain spectra of the plurality of different processes performed by one or more machines of the first type; and / or 7. The method of claim 6, wherein the filter frequency domain spectrum of the reference machine is based on an average of the set of frequency domain spectra of the plurality of different processes performed by one or more machines of the second type.

8. 8. The method of claim 7, wherein the new frequency domain spectrum of the process performed by the new machine is one of the set of frequency domain spectra of processes performed by the one or more machines of the first type.

9. The method of claim 1 further comprising interpolating the harmonized new machine specific weights across frequencies.

10. The method of claim 1 , wherein the new frequency domain spectrum is received from the new machine or from local storage.

11. receiving time series data of the process performed by the new machine; and 2. The method of claim 1, further comprising transforming the time series data into the new frequency domain spectrum of the process performed by the new machine.

12. 12. The method of claim 11, further comprising recovering time-domain data from the harmonized new frequency-domain spectrum.

13. The method of claim 1 , wherein the data is electroencephalogram data.

14. The method of claim 1 , wherein the method is a computer-implemented method.

15. 1. A method for generating a set of harmonized new machine-specific weights for harmonizing data from a new machine of a first type with data from a reference machine of a second type, the method comprising: The method includes determining a set of harmonized new machine specific weights based on a filter frequency domain spectrum of a hardware filter of the reference machine and a filter frequency domain spectrum of a hardware filter of the new machine.

16. receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the first type; and 16. The method of claim 15, further comprising determining the filter frequency domain spectrum of the hardware filter of the new machine based on the set of frequency domain spectra of the plurality of different processes performed by the one or more machines of the first type.

17. receiving a set of frequency domain spectra of a plurality of different processes performed by one or more machines of the second type; and 16. The method of claim 15, further comprising determining the filter frequency domain spectrum of the hardware filter of the reference machine based on the set of frequency domain spectra of the plurality of different processes performed by the one or more machines of the second type.

18. 18. The method of any one of claims 15 to 17, wherein the method is a computer-implemented method.

19. 1. A device for reconciling data from a new machine of a first type with data from a reference machine of a second type, comprising: the device includes a processor and a memory; The memory includes machine-executable instructions that, when executed on the processor, cause the processor to: receiving a new frequency domain spectrum of a process performed by the new machine; and determining a harmonized new frequency domain spectrum of the process performed by the new machine; the harmonized new frequency domain spectrum is harmonized with a corresponding frequency domain spectrum of the process when the process is performed by the reference machine by applying a set of harmonized new machine-specific weights to the new frequency domain spectrum; The set of harmonized new machine specific weights is determined based on a filter frequency domain spectrum of a hardware filter of the new machine and a filter frequency domain spectrum of a hardware filter of the reference machine.

20. 20. The device of claim 19, wherein the memory includes machine-executable instructions that, when executed on the processor, cause the processor to perform the method of claim 14.

Images