WO2017108864A1

WO2017108864A1 - A system and a method for estimation of arterial blood gas

Info

Publication number: WO2017108864A1
Application number: PCT/EP2016/082034
Authority: WO
Inventors: Ravindra Balasaheb PATIL; Krishnamoorthy PALANISAMY; Rithesh SREENIVASAN; Meru Adagouda PATIL
Original assignee: Koninklijke Philips N.V.
Priority date: 2015-12-21
Filing date: 2016-12-20
Publication date: 2017-06-29

Abstract

A system for estimation of Arterial Blood Gas (ABG). The system comprises an ultrasound element to provide Doppler signal to identify the artery having blood flow, from which ABG is to be estimated. The system further comprises an InfraRed (IR) element to provide IR signal having IR data required for estimation of ABG, and a signal unit for transmitting the Doppler signal and the IR signal, and receiving the reflected Doppler signal and the reflected IR signal. The system has a processing unit to perform localization of the artery and to provide a model for estimation of ABG based on the IR data. There is also a method for estimation of Arterial Blood Gas (ABG) by the system.

Description

A SYSTEM AND A METHOD FOR ESTIMATION OF ARTERIAL BLOOD GAS

TECHNICAL FIELD

[0001] The present disclosure relates to estimation of Arterial Blood Gas (ABG), more particularly to non-invasive and direct estimation of ABG.

BACKGROUND

[0002] The measurement of Arterial Blood Gas indicates the levels of oxygen and carbon dioxide in the blood from an artery. This measurement provides indication on the capability of the lungs to move oxygen into the blood and remove carbon dioxide from the blood. Apparently, the measurement of Arterial Blood Gas includes the measurement of arterial oxygen tension (Pa0₂), carbon dioxide tension (PaC0₂), acidity (pH) along with other parameters like oxygen saturation and bicarbonates. These measurements become vital for subjects with critical illness or respiratory disease. Hence for these reasons, Arterial Blood Gas is considered to be one of the most common and important tests that needs to be performed for critical care patients. This test is routinely performed in Intensive Care Unit (ICU) setting. Besides ICUs, it is extensively used in the diagnosis of heart failure, kidney failure, uncontrolled diabetes, sleep disorders, severe infections, or after a drug overdose etc.

[0003] Currently, the Arterial Blood Gas is more reliably estimated by invasive techniques. It involves puncturing of an artery with a needle or syringe to draw blood. The puncturing of the artery is done either at the site of radial artery or femoral artery. This technique cannot be applied in continuous monitoring as the test by itself is non-continuous. Moreover, this technique needs skilled care-giver as well as it causes pain and discomfort to the patients. Besides this, the results cannot be determined instantly or faster and hence time consuming.

[0004] Alternately, Arterial Blood Gas can also be measured through non-invasive techniques or approaches. One such approach is disclosed in US 5632281, in which the volume and carbon dioxide concentration of the expiratory breath are measured and breath volumetric rate and gas content are discerned therefrom. Further processing is performed to derive Arterial Blood Gas levels of carbon dioxide. It is to be understood that the estimation of Arterial Blood Gas from exhaled air is not comparable to the estimation of Arterial Blood Gas from blood. Also, this approach requires the patient to exhale forcefully and may not be possible when the patient is administered with anesthesia or with the neonates or unconscious patients. Moreover, this is not a continuous approach and hence cannot be applied in continuous monitoring and also requires the device or apparatus deployed to be calibrated from time to time.

[0005] Owing to the above short comings, there is a need for a solution to measure ABG non-invasively and could be more reliably used for diagnosis.

SUMMARY

[0006] This specification is directed towards a system for estimation of Arterial Blood Gas (ABG). The system includes in various aspects an ultrasound element to provide Doppler signals which may be used to identify the artery having blood flow, from which ABG is to be estimated. The system further may include in some aspects an Infra-Red (IR) element to provide IR signal having IR data required for estimation of Arterial Blood Gas, and a signal unit for transmitting the Doppler signal and the IR signal, and receiving the reflected Doppler signal and the reflected IR signal. The system may further include a processing unit to perform localization of the artery and to provide a model for estimation of Arterial Blood Gas based on the IR data.

[0007] The present disclosure further provides a method for estimation of Arterial Blood Gas (ABG) by the system in accordance with the disclosure. The method provided includes various steps of conditioning the Doppler signal provided by the ultrasound element and the IR signal provided by the IR element, performing localization of the artery having blood flow from which ABG is to be estimated by a processing unit. The method of the present disclosure may further include generating a model for estimation of Arterial Blood Gas from the IR data of the reflected IR signal, and estimating the Arterial Blood Gas through estimation of Pa0₂ and PaC0₂ parameters of the blood by the model.

[0008] In various aspects, the present disclosure provides a system for estimation of ABG through non-invasive approach. Such a system may include estimation of ABG that is more reliable and provide estimation of ABG in a faster and continuous real time manner. The present disclosure further provides a method for estimation of ABG using the various aspects and features of the apparatus and system disclosed herein.

[0009] In various aspects, the present disclosure sets forth a method for estimation of arterial blood gas, comprising the steps of: conditioning the Doppler signal provided by the ultrasound element and the IR signal provided by the IR element; performing localization of the artery having blood flow from which ABG is to be estimated, by a processing unit; generating over a trained model, estimation of arterial blood gas from the IR data of the reflected IR signal; and estimating of the arterial blood gas through estimation of Pa0₂ and PaC0₂ parameters of the blood by the trained model.

[0010] This method and other implementations of technology disclosed herein may each optionally include one or more of the following features.

[0011] In some implementations, the method further includes performing localization of the artery by identifying the artery based on the Doppler signal provided by the ultrasound element, the ultrasound element operating in duplex mode. In some versions, the method of identifying the artery further includes detecting the flow of the blood in the blood vessel based on the reflected Doppler signal of the ultrasound element. In some other versions, the identifying the artery step further includes identifying the blood vessel as an artery or a vein. In still other versions, identifying the artery also includes providing feedback on the identified blood vessel. In yet other aspects, the present method includes estimating the arterial blood gas directly and / or continuously in real time or offline. In yet other versions, the method includes estimating the arterial blood gas non-invasively.

[0012] In some implementations, the disclosure herein may include a computer program product which incorporates instructions, performed on a processor and stored in memory, the steps outlined in the method above when executed on a computer. Still other aspects may include a computer readable medium having the above mentioned computer program product.

[0013] In some aspects and embodiments, a system may be provide for estimation of a rterial blood gas, which has, among other features, an ultrasound element to provide a Doppler signal to identify an artery having blood flow; an Infra Red (IR) element to provide an IR signal having I R data for estimation of ABG; a signal unit for transmitting the Doppler signal and the IR signal, and receiving the reflected Doppler signal and the reflected IR signal; and a processing unit having at least one processor configured to perform localization of the artery and to provide a model for estimation of ABG based on the IR data.

[0014] This method and other implementations of technology disclosed herein may each optionally include one or more of the following features.

[0015] In various implementations, the system may further include an ultrasound element is a single ultrasound element integrated with an IR transmitter and receiver. In other aspects, the system may further include transmission of the wavelength of the I R signals in the range of 700 nm to lOOOnm. Still other versions may have a signal unit which is provided to condition the Doppler signals and the IR signals. Some versions may also include a signal unit provided to vary the wavelength of the IR signal. In some additional versions, the processing unit is configured to perform a fourier transform on the IR data of the reflected IR signal. Further aspects of the system may include a model which is a regression model. Still additional embodiments of the system may include an ultrasound element provided to identify the artery based on a non-imaging approach and which operates in duplex mode.

[0016] In some implementations, the model may be provided for estimation of arterial blood gas through estimation of Pa0₂ and PaC0₂ parameters of the blood, either directly or continuously. In other versions, the system may provide such estimation in real time or offline. Still other aspects of the system is to provide estimation of arterial blood gas non- invasively.

[0017] Other implementations may include a non-transitory computer readable storage medium storing instructions executable by a processor (e.g., a central processing unit (CPU) or graphics processing unit (GPU)) to perform a method such as one or more of the methods described above. Yet another implementation may include a system of one or more computers and/or one or more devices that include one or more processors operable to execute stored instructions to perform a method such as one or more of the methods described herein.

[0018] In other aspects, the present disclosure sets forth a method of training a model, possibly a neural network regression model for estimation of arterial blood gases. In some aspects, the method of training includes identifying, by one or more processors, a plurality of training examples generated based on IR signal from one or more IR elements reading reflected IR from a patient arterial position. Other features may include training examples each including training example input having reflected IR from an emitted IR ranging from a wavelength of 700nm to 1000 nm; determined fast Fourier transforms and power spectral densities of the reflected IR; determined wavelet transforms of the reflected IR; feature set extractions to form feature vectors. Still further, each of the training examples may include training example output having a Pa02 and PaC02 value in mmHg, wherein the training, by one or more of the processors, creates regression model based on the training examples, the training comprising iteratively updating the neural network regression model based on application of the training example input and the training example output of each of the training examples to the neural network regression model.

[0019] It should be appreciated that all combinations of the foregoing concepts and

additional concepts described in greater detail herein are contemplated as being part of the subject matter disclosed herein. For example, all combinations of claimed subject matter appearing at the end of this disclosure are contemplated as being part of the subject matter disclosed herein.

BRIEF DESCRIPTION OF THE DRAWINGS [0020] With reference to the accompanying drawings in which: [0021] Figure 1 shows a system for estimation of ABG in accordance with the description herein.

[0022] Figure 2 shows a method for estimation of ABG in accordance with the description of the system set forth herein.

[0023] Figure 3 schematically illustrates an exemplary environment for utilizing a system for estimation of Arterial Blood Gas as described herein.

[0024] Figure 4 illustrates an example flowchart of estimating ABG in various implementations described herein.

[0025] Figure 5 illustrates an example flowchart for the reflected IR quantification before submitting to the trained machine learning model as is described herein.

DETAILED DESCRIPTION

[0026] Various implementations of the disclosure provided herein set forth a system and method for estimation of Arterial Blood Gas. Such system and method enable estimation of the ABG without utilization of the invasive techniques previously noted by implementation of an ultrasound element and an infrared element on combination with a processing unit to detect and process the signals from signal unit. In some implementations, the system will utilize the ultrasound element to appropriately locate and identify the arterial or vein location and mark such location. Such implementations may further utilize an infrared element obtain in real time reflected IR signals which may be analyzed and based on a model, estimate concentrations of Pa02 and PaC02 in mmHG. The system and method is further described hereinafter with reference to non-exhaustive exemplary embodiments and with reference to the Figures. Additional description of these and other implementations of the technology are described below.

[0027] Figures 1 and 2 show a system and a method, respectively for estimation of Arterial Blood Gas (ABG). The system (100) for estimation of ABG comprises of a single ultrasound element (101), infra-red element (102), a signal unit (103) and a processing unit (104). [0028] The ultrasound element (101) provides a Doppler signal to identify the artery having blood flow from which the ABG is to be estimated. The Doppler signal is provided by the ultrasound element (101) and is transmitted by the signal unit (103) after conditioning (201) of the Doppler signal. The identification of the artery is performed through localization of the artery (202) by the processing unit (104) based on the Doppler reflected from the blood vessel reflecting it.

[0029] As outlined in the various implementations and in particular in the aspects of Figure 4, the system presented and described herein may include, using a combined device for ultrasound and IR transmission and reception, Doppler signal acquisition of the prospective patient site at step 402. The system and method may then, at step 404, extract the spectral profile of the data obtained through the Doppler acquisition. At step 406, the system may estimate the turbulence for the proposed site in order to properly identify, at step 408 a vein or arterial location. If such an artery is localized at stop 410, the region, in various aspects and implementations, may be appropriately marked at step 412. Once the system has appropriately determined the location of the artery, a the IR element may be activated at step 414 such that the reflected IR may be received and appropriately quantified at step 416. After quantification, the IR signal data may be fed into the trained machine learning model at step 420 in order to estimate the Pa02 and PaC02 values at step 418.

[0030] In various aspects, the device described herein may include a single housing which may include both the ultrasound element and the IR element as well as respective signal unit processing and other electronics necessary for gating, transmitting and receiving the appropriate signals, as well, the device may further include hepatic feedback features, readouts or other user interface aspects and communication devices for receipt and transmission of data and/or signals to other structural elements, such as controllers, processing units and the like. Additionally, in various embodiments, the system and method set forth herein may include a plurality of processors each configured to implement various steps and aspects outlined herein. Such system may be integrated or separated, each of the individual elements being operable to control and/or communication necessary signals and/or data to another connected element and or portion. As well, the trained learning model may as well be integrated into memory and/or storage associated with the unitary device, it may also be associated with the caregiver portable device, handheld computer or separate system.

[0031] Although the methods outlined herein and system are illustrated in separate figures for the sake of clarity, it is understood that one or more blocks of method may be performed by the same component(s) that perform one or more blocks of the method. For example, one or more (e.g., all) of the blocks of method and the system may be performed by processor(s) of a single handheld device or may be segmented into separate devices as may be warranted. Also, it is understood that one or more blocks of the method and system may be performed in combination with, or preceding or following, one or more blocks of method.

[0032] Many of the above mentioned steps may also be implemented by one or more processors which have associated memory with instructions, the processor configured to complete each of the steps thereby applying the various treated IR signal data to the trained model and or aid in training the model using a plurality of training examples.

[0033] As is further depicted in Figure 3, the processing unit 104 is operably in communication with the signal unit 103 which may, in various implementations, control transmission and receipt of the ultrasound transducers or other electronic components found at the combined ultrasound element and IR element 101/102. The combined ultrasound and IR element 101/102 can further operate in duplex mode in order to acquire data from the patient, as shown, in order to detect and localize the artery 202. Thus, the Doppler signals are detected to identify blood flow and the signals themselves are segregated into either arterial flow or vein so that a proper arterial flow area is detected, identified and marked.

[0034] In various embodiments, the ultrasound element may include a transducer array containing a plurality of elements that each produce ultrasonic energy when energized. The system 100 is designed to aid in positioning of the IR element 102 within an arterial area of interest, such as being intravascularly positioned along an axis of an arterial blood vessel. In various embodiments, the transducer array may be designed to emit ultrasound signals directed toward the tissue area of interest in response to which Doppier signals are detected. Such receipt of Doppier echoes from patient target areas which are transverse to the axis of the device 101/102 may be associated with a desirable arterial blood vessel. The Doppier signals, via the rebounded ultrasonic energy, includes information pertaining to the patient target areas and is converted to an electrical signal by various transducers, for example, of the ultrasound element. Such signals may then be provided to a receiver either configured internally within the transducers or integrated within the ultrasound element. The ultrasound element transmitter, receiver and various supporting electronics of the ultrasound element 101 are operated under the control of a controller or combined controller, signal processor unit, processing unit or the like. A scan is performed by acquiring a plurality of rebounded echoes in which the element 101 is set to intermittent transmit and receipt conditions, such that it is able to both transmit and receive subsequent rebounded signals. Separate rebound signals from each transducer element or other receiver electronics are then combined to produce a single ultrasound signal which may then be analyzed or modified in the signal unit 103.

[0035] Localization of an artery, in various embodiments, includes detecting the flow of blood in the blood vessel and identifying whether the blood vessel is an artery or a vein by segregating the rebounded/reflected Doppier signal in respect of the signal reflected from the artery or vein. Upon identification of the blood vessel 202 being an artery, the localization of the artery or its type is performed. Localization of the artery is based on non-imaging approach of the Doppier ultrasound. A single element ultrasound element with RF signal, with no beam forming, is sufficient to identify the artery.

[0036] The identification and localization of artery is performed by the processing unit (104) based on the reflected Doppier signal, in various steps as shown in Figure 5 and noted below.

[0037] At step 501, a Fast Fourier Transform (FFT) is performed for every frame of 12.8 ms (Bi),

H_t = FFT(Bi) [0038] At step 502, the FFT output is conditioned by High pass filter with cut-off frequency of 500 Hz

H i(fc) = FFT(Hi), k = 1 N, (N=Length of the frame)

[0039] At step 503, the cu mulative sum of energy is computed as follows

CS(n) =∑l₌₁ \HF_i(k) \, n=l N

[0040] At step 504, the point where the cu mulative energy crosses 50% of total energy is determined by: (0 =r, such that ∑^r _n=1 C5(n) = 0.S∑ ₌₁ CS(n

[0041] The above steps are repeated until the 5 seconds of data is completed.

[0042] The ensemble of these points from each frame constitutes a feature vector. At step 505, the standard deviation of feature vector is computed, and at step 506, the signal with standard deviation greater than predetermined threshold is selected.

[0043] At step 507, the spectrogram is obtained using the Fast Fourier Transform (FFT) based power spectrum with a 20-millisecond Hamming window using 256 data points and 50% overlap. A smooth and reproducible maximum frequency envelope is extracted from the spectrogram using 2D Teager operator.

[0044] At step 508, a higher order statistic parameter is computed on the extracted envelope e(t) = V*( ² + x^A(t)²

where, x is signal envelope and x^A(t) is the Hilbert transform of x(t).

[0045] At step 509, a signal is classified as an artery if HOSP (Higher Order Statistical Power) is greater than predefined threshold .

[0046] Upon identification and localization of the artery, a tactile feedback is provided to the signal unit (103) which deactivates the ultrasound element (101). The region on the skin where the localization of the artery is made is marked by a marker (not shown). The signal unit (103) enables the transmission of Infra-Red (IR) signal from the Infra-Red (IR) element (102) onto the marked region where the localization or identification of the artery is done.

[0047] In various embodiments, the signal unit (103) may be provided for transmitting the Doppler signal by the ultrasound element (101) and of the IR signal by the IR element (102). The signal unit (103) may also receive the Doppler signal and IR signal reflected from the blood vessel. The signal unit (103) may further, in alternative aspects, condition the Doppler signal and the IR signal and of the reflected signals thereof.

[0048] A model (105) is generated (203) for estimation of ABG (204) based on the IR data from the reflected IR signal. The estimation of the ABG is based on the detected measurement of Pa0₂ and PaC0₂.

[0049] In real time phase the reflected IR signal is analyzed and based on the model, concentration of Pa02 and PaC02 in mmHG is estimated. The direct measurement of Pa02 and PaC02 by analyzing the IR spectra using non-invasive approach at a single site after identification of the artery is made possible.

[0050] Upon the identification of the artery, the spectra of IR ranging from wavelength 700 nm to 1000 nm with 10 sub-bands is impinged on the marked region on the skin. This variation of IR wavelength is carried out by the signal unit (103) or in alternative embodiments by associated electronics which are available, to appropriately drive the IR element and Ultrasonic element. The reflected IR signal is captured from the skin surface stored or transmitted for handling and/or analysis. Fourier Transform (FT) of the reflected IR signal is performed succeeding which power spectral analysis is computed. The coefficient of the FT, along with the following frequency domain features, are extracted from the PSD (Power Spectral Density) obtained from the non-parametric method: (i) Variance, (ii) Skewness, (iii) Kurtosis, (iv) 3 dB bandwidth, (v) Spectral entropy, (vi) Average peak frequency, and (vii) 95% spectral roll-off. 1] The spectral centroid indicates the position of the centre of mass of the spectrum and is calculated as the weighted mean of the frequencies present in the signal, determined using the PSD, with their magnitudes as the weights.

That is,

_∑ _{= 1} k PSD(k)

∑ ₌₁ PSD(k) ·

[0052] Here, PSD(k) is the amplitude corresponding to bin k in PSD.

[0053] The average peak frequency is defined in some embodiments and implementations as the frequency around which maximum power is present. To calculate this frequency, the PSD is low-pass filtered with an order of 'M'. That is,

YW = b_kPSD(n - K); b_k =

[0054] The frequency at which Y(k) is maximum is selected as average peak frequency.

M is the filter order

[0055] The 95% roll off frequency for a given PSD is defined as the frequency below which 95% of the power of the signals is contained. Mathematically, the 95% roll off frequency is

Kl, X^ PSDVc] = 0.95 X∑^₌₁ PSD(k)

where N corresponds to frequency index of maximum frequency component (i.e., number of FFT points).

[0056] Also, discrete wavelet transform is performed on the signal to extract the wavelet coefficients. The extracted wavelet coefficients provide a compact representation that shows the energy distribution of the signal in time and frequency, the wavelet used with three level decomposition.

[0057] Let, CAi and CDi are the approximation and detail coefficients of level Ύ, respectively, Li is the length of CDi, and N is the number of decomposition levels. The following parameters from the coefficients are defined.

y CD?

[0058] Power spectral density: PD_j = -— ^L Absolute mean: AM; =

Li Average amplitude change:

_∑|CD_i(n) - CD_i(n - l) |

A^Li - Li

Difference absolute standard deviation:

Fractal length:

From these parameters, the following wavelet features are extracted Ratio of power spectral density:

PDi-i

Ratio of absolute mean:

AM_i_₁

z AMi

Ratio of average amplitude change:

Ratio of difference absolute standard deviation:

Difference in power spectral densities:

WF₅ (Q = PD_i_₁ - PD_i

Wavelet power ratio:

Power variation in the detail coefficients:

Ratio of fractal length: ^WF°W = -TET

[0059] All the above mentioned feature set forms a feature vector to generate the regression model for estimating PaC02 and Pa02. After the model is trained, the value of PaC02 and Pa02 in mmHg is estimated for the new feature inputs directly and noninvasive^ in real time. Some implementations set forth herein are directed towards training a model, possibly programmer defined algorithms, Bayesian networks, neural networks or deep neural network, etc., to predict and estimate the value of PaC02 and Pa02 in mmHg based upon reflected IR over a determined localized arterial site. Other implementations enable entering into a trained model, after localization of an Artery, quantification of the reflected IR in order to estimate the PaC02 and Pa02. Still further implementations are directed towards utilization of a trained model to provide an estimate of ABG through non-invasive means. For example, a trained neural network may be utilized in the iterative updating of a model for more accurate estimations of the ABG by entering into the trained model reflected or modified data signals representative of the reflected IR signals.

[0060] Thus, in various implementations, a method of training a neural network regression model or other type of model for estimation of arterial blood gases is provided. Such a method may include, in aspects, identifying, by one or more processors, a plurality of training examples generated based on IR signal from one or more IR elements reading reflected IR from a patient arterial position. The training examples may each include training example input having relevant data such as reflected IR from an emitted IR ranging from a wavelength of 700nm to 1000 nm; determined fast Fourier transforms and power spectral densities of the reflected IR; determined wavelet transforms of the reflected IR; and feature set extractions to form feature vectors. Such training examples may be framed from the various implementation methods outlined herein and defining the step of estimation for the a02 and PaC02. As well, the training examples may further include training example output, the output having an data identifier representative of the actual Pa02 and PaC02 value in mmHg. In such a training of a regression model, the training may take iteratively such that the regression neural network is updated based upon application of the training example input and the training example output of each of the training examples being applied to the neural network. In implementations, the training may include performing backpropagation on the network based on the training example output of the plurality of training examples.

1] While several implementations have been described and illustrated herein, a variety of other means and/or structures for performing the function and/or obtaining the results and/or one or more of the advantages described herein may be utilized, and each of such variations and/or modifications is deemed to be within the scope of the implementations described herein. More generally, all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and/or configurations will depend upon the specific application or applications for which the teachings is/are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific implementations described herein. It is, therefore, to be understood that the foregoing implementations are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, implementations may be practiced otherwise than as specifically described and claimed. Implementations of the present disclosure are directed to each individual feature, system, article, material, kit, and/or method described herein. In addition, any combination of two or more such features, systems, articles, materials, kits, and/or methods, if such features, systems, articles, materials, kits, and/or methods are not mutually inconsistent, is included within the scope of the present disclosure.

Claims

WE CLAIM:

1. A system for estimation of Arterial Blood Gas, comprising:

an ultrasound element to provide a Doppler signal to identify an artery having blood flow;

an Infra Red (IR) element to provide an IR signal having IR data for estimation of ABG; a signal unit for transmitting the Doppler signal and the IR signal, and receiving the reflected Doppler signal and the reflected IR signal; and

a processing unit having at least one processor configured to perform localization of the artery and to provide a model for estimation of arterial blood gas based on the IR data;

the model receiving a feature vector representing quantification of the reflected IR and outputting an estimate of PaC02 and Pa02.

2. The system as claimed in claim 1 wherein the ultrasound element is a single ultrasound element.

3. The system as claimed in claim 1 wherein the wavelength of the IR signal is in the range of 700 nm to lOOOnm.

4. The system as claimed in claim 1 wherein the signal unit is provided to condition the Doppler signal and the IR signal.

5. The system as claimed in claim 1 wherein the signal unit is provided to vary the wavelength of the IR signal.

6. The system as claimed in claim 1 wherein the processing unit is configured to perform a fourier transform on the IR data of the reflected IR signal.

7. The system as claimed in claim 1 wherein the model is a regression model.

8. The system as claimed in claim 1 wherein the ultrasound element is provided to identify the artery based on a non-imaging approach and operates in duplex mode.

9. The system as claimed in claim 1 wherein the model is provided for estimation of arterial blood gas through estimation of Pa0₂ and PaC0₂ parameters of the blood.

10. The system as claimed in any one of the preceding claims, wherein the system is provided for estimation of arterial blood gas directly and / or continuously in real time or offline.

11. The system as claimed in any one of the preceding claims, wherein the system is provided for estimation of arterial blood gas non-invasively.

12. A method for estimation of Arterial Blood Gas (ABG), comprising the steps of:

conditioning the Doppler signal provided by the ultrasound element and the IR signal provided by the IR element;

performing localization of the artery having blood flow from which ABG is to be estimated, by a processing unit;

determining a feature set to form a feature vector from reflected IR data of the IR signal;

generating over a trained model, estimation of arterial blood gas from the reflected IR data based upon the feature vector; and

the estimating of the arterial blood gas obtained through estimation of Pa0₂ and PaC0₂ parameters of the blood by the trained model.

13. The method as claimed in claim 12 wherein performing localization of the artery includes identifying the artery based on the Doppler signal provided by the ultrasound element, the ultrasound element operating in duplex mode.

14. The method as claimed in claim 13 wherein identifying the artery includes detecting the flow of the blood in the blood vessel based on the reflected Doppler signal of the ultrasound element.

15. The method as claimed in claims 13 wherein identifying the artery includes identifying the blood vessel as an artery or a vein.

16. The method as claimed in claims 13 wherein identifying the artery includes providing feedback on the identified blood vessel.

17. The method as claimed in any one of the claims 12 further including estimating the arterial blood gas directly and / or continuously in real time or offline.

18. The method as claimed in any one of the claims 12 further including estimating the arterial blood gas non-invasively.

19. A computer program product comprising code means for performing the method as claimed in claims 12 to 18 when executed on a computer.

20. A computer readable medium comprising the computer program product as claimed in claim 19.

21. A method of training a neural network regression model for estimation of arterial blood gases, comprising:

identifying, by one or more processors, a plurality of training examples generated based on IR signal from one or more IR elements reading reflected IR from a patient arterial position, each of the training examples including training example input comprising:

reflected IR from an emitted IR ranging from a wavelength of 700nm to 1000 nm; determined fast Fourier transforms and power spectral densities of the reflected IR;

determined wavelet transforms of the reflected IR;

feature set extractions to form feature vectors;

each of the training examples including training example output comprising: a Pa02 and PaC02 value in mmHg;

training, by one or more of the processors, a regression model based on the training examples, the training comprising iteratively updating the neural network regression model based on application of the training example input and the training example output of each of the training examples to the neural network regression model.