Systems and methods for determining blood pressure of an individual

Abstract

A method for measuring blood pressure values of an individual includes receiving video data of one or more regions of interest (ROIs) of the individual’s face 110. At least one colour channel of the c

Description

TECHNICAL FIELD The present disclosure relates to a system and method for determining physiological parameters of an individual, and particularly to a system and method for (remote) measurement, assessment and monitoring of parameters such as blood pressure, heart rate and respiratory rate. More specifically, the system and method utilise video data of a portion of the individual’s skin, and analyse the video data to extract details of the variations in light passing through the skin in order to determine the physiological parameters. BACKGROUND There have been many developments, over the years, in the healthcare sector and in relation to the monitoring of an individual’s health. One mechanism for monitoring the health of an individual involves the measurement (and assessment over time) of the physiological parameters of the individual. Such parameters include the commonly-measured vital signs of an individual such as their systolic and diastolic blood pressure (BP); respiratory rate (RR); heart rate (HR); and heart rate variability (HRV). Such parameters have traditionally been measured in person, for example, by a medical practitioner, during a physical medical check-up of the individual. Developments in recent years have increasingly focused on the ability of an individual to measure and monitor their vital signs in their own homes (without needing the physical presence of a medical practitioner). For example, the use of wearable technology that measures and monitors some of these parameters (e.g., heart rate) is becoming increasingly widespread and easily accessible to individuals to monitor their vital signs on their own. Further developments are also ongoing that would allow the individual to measure at least some of their vital in a remote manner (without necessarily relying on the use of wearable technology). Nevertheless, even in instances where vital signs can be measured and monitored by the individual’s themselves, any subsequent clinical recommendations I diagnoses would involve analysis of the vital signs data by a medical practitioner. Such analyses would still be subject to a degree of variability since they depend on the medical practitioner’s subjective interpretation of the data. It is against this background that the present disclosure has been set. SUMMARY OF THE DISCLOSURE According to an aspect of the present disclosure, there is provided a computer implemented method for measuring blood pressure values of an individual. The method includes receiving video data of one or more regions of interest ROIs of the individual’s face, and processing at least one colour channel of the captured video data to identify and obtain a time-varying photoplethysmography PPG signal. The PPG signal is provided as an input to at least one neural network to derive a representation of a signal waveform corresponding to the arterial blood pressure ABP of the individual, one or more blood pressure values of the individual are determined from the ABP signal waveform and the determined one or more blood pressure values are outputted to enable the blood pressure of the individual to be monitored. The above-described method provides the ability to derive the individual’s blood pressure BP directly from the video data (via use of a determined PPG signal and one or more neural networks). This reduces the amount of intermediate processing that is required (and the potential errors and approximations that might be associated therewith) in order to derive the blood pressure, ensuring a relatively easy mechanism for ascertaining BP without increasing the associated errors. It is useful to extract data from these ROIs of the user’s face, as it reduces the volume of data that needs to be processed thereafter to determine blood pressure values of the individual, relative to if the entire image was to be processed. Furthermore, by capturing video data, various characteristics and features of the user can be detected such as the age, sex I gender, ethnic origin group of the user, and therefore, the implementation of medicine can be tailored to best suit the individual. Optionally, determining one or more blood pressure values can include at least one of: (i) determining a systolic blood pressure SBP corresponding to a maximum of the ABP signal waveform, (ii) determining a diastolic blood pressure DBP corresponding to a minimum of the ABP signal waveform, or (iii) determining a mean arterial pressure MAP corresponding to an average of the ABP signal waveform. In some embodiments, processing at least one colour channel of the captured video data includes processing a plurality of colour channels including at least red, blue and green colour channels, filtering the PPG signal in each of the plurality of colour channels to attenuate irregularities and noise sources and benchmarking the filtered PPG signal in each of the plurality of colour channels to select the best channel for subsequent analysis. Benchmarking is beneficial as it takes various different factors I conditions / properties into account that affect the quality of the resulting signal such as the pigment of the individual’s sign and the conditions of the surrounding environment and can select the most appropriate signal for further analysis. In some instances, the at least one neural network corresponds to one or more convolutional neural networks CNNs that have previously been trained by machine learning to derive the ABP signal waveform from the input PPG signal. In such instances, the method can further include a prior process of training the one or more CNNs. The training process includes programming the one or more CNNs with an algorithm defining an initial relationship between PPG signal and ABP signal waveform and (iteratively) refining the algorithm of the one or more CNNs. The algorithm is refined by applying the one or more CNNs to training data sets of video data obtained from a set of users to derive the BP of each user in the training data, comparing the derived BP with ground truth values of the respective user’s BP measured previously using an alternative mechanism and regressing the estimated vital signs against the corresponding ground truth signs. By refining the algorithm, the ABP signal waveform becomes closely approximate to the actual ABP of the individual and analysis of this resultant waveform can be carried out in order to determine key BP parameters. In another embodiment, the at least one neural network includes a first neural network and a second neural network, and wherein providing the PPG signal to the at least one neural network includes providing the PPG signal to the first neural network to obtain first approximation of the ABP signal waveform and providing the approximated ABP signal waveform to the second neural network to refine the signal waveform and obtain the final representation of the ABP signal waveform. In some embodiments, receiving video data of at least a portion of the individual’s face includes establishing a video communications channel between the individual and a medical practitioner and capturing the video data during the video communication. In such instances, outputting the determined value of the at least one physiological parameter may include outputting the determined value of the at least one physiological parameter to at least one of: a medical practitioner and / or the individual in real time during the video communication. Outputting the at least one medical conclusion can include outputting, to the individual and / or a medical practitioner, a message relating to an action to be taken by the individual and / or medical practitioner based on the at least one medical conclusion output. The video communications channel can be used for remote medical consultations which may enhance access and bring other important benefits to a patient such as reduced travelling and greater flexibility. Outputting the conclusion to the medical practitioner is useful as it is straightforward for the medical practitioner to provide advice to the subject, and it is also easy for the subject to be provided with these instructions. In some instances, outputting the determined one or more blood pressure values further includes storing the determined one or more blood pressure values in association with an identifier uniquely identifying the individual. In another embodiment, the method further includes retrieving data relating to one or more individualspecific properties of the individual, analysing the determined one or more blood pressure values based on the retrieved data and a set of predefined rules to determine a medical conclusion regarding hypertension, and outputting the determined medical conclusion. In such instances, the method can further include determining the one or more blood pressure values multiple times over a period of time to generate a historical set of blood pressure values, storing the historical set of blood pressure values in association with the unique identifier. Analysing the determined one or more blood pressure values includes retrieving the historical set of values for analysis in combination with the determined one or more blood pressure values based on the retrieved individual-specific data and the set of predefined rules to determine the medical conclusion. Optionally, the retrieved individual-specific data can include one or more of the following: an age of the individual; a gender of the individual; a weight of the individual; a demographic profile of the individual; or a socioeconomic profile of the individual. When taken in combination with the medical conclusion, these features allow the conclusion itself to be contextualised and personalised; it is hence more accurate and useful. In some embodiments, the predefined set of rules is based on clinically approved assessment guidelines, optionally comprising at least one of: BM J Best Practice guidelines. In some instances, the method further includes transmitting the determined one or more blood pressure values to a remote healthcare provider system for storage and / or ongoing monitoring of the individual. This essentially creates a library for each individual where the data in the library can be subsequently monitored, for example, when using historical data and contextual data based to assess the progress of a (suspected) chronic disease. Some benefits of this include remote identification of acute exacerbations of an individual’s chronic condition, allowing contextual understanding of changes in disease progress, ensuring general health monitoring, and ensuring any adverse drug reactions, such as an allergy or an interaction, would be identified as early as possible. The benefits obtainable via this digital monitoring could improve confidence in remote prescribing and better care decisions. Optionally, the method further includes determining a value of one or more additional physiological parameters of the individual from the PPG signal, the one or more additional physiological parameters comprising at least one of: heart rate (HR); respiratory rate (RR); and heart rate variability (HRV). According to another aspect of the present disclosure, there is provided a system for measuring blood pressure values of an individual, the system comprising: an input configured to receive video data of one or more regions of interest ROIs of the individual’s face; at least one processor configured to process at least one colour channel of the captured video data to identify and obtain a time-varying photoplethysmography PPG signal; provide the PPG signal as an input to at least one neural network to derive a representation of a signal waveform corresponding to the arterial blood pressure ABP of the individual; and determine one or more blood pressure values of the individual from the ABP signal waveform. The system further comprises an output configured to output the determined one or more blood pressure values to enable the blood pressure of the individual to be monitored. Features and benefits that were associated with the above-described method are also equally applicable to the corresponding system. For instance, the at least one processor may be configured to: (i) determine a systolic blood pressure SBP corresponding to a maximum of the ABP signal waveform, (ii) determine a diastolic blood pressure DBP corresponding to a minimum of the ABP signal waveform, or (iii) determine a mean arterial pressure MAP corresponding to an average of the ABP signal waveform. In some embodiments, the at least one processor may be configured to process a plurality of colour channels including at least red, blue and green colour channels, filter the PPG signal in each of the plurality of colour channels to attenuate irregularities and noise sources and benchmark the filtered PPG signal in each of the plurality of colour channels to select the best channel for subsequent analysis. In some instances, the at least one neural network corresponds to one or more convolutional neural networks CNNs that have previously been trained by machine learning to derive the ABP signal waveform from the input PPG signal. In such instances, the at least one processor may be configured to implement a prior process of training the one or more CNNs. In the training process, the at least one processor is configured to program the one or more CNNs with an algorithm that defines an initial relationship between PPG signal and ABP signal waveform and to (iteratively) refine the algorithm of the one or more CNNs. To refine the algorithm, the at least one processor is configured to apply the one or more CNNs to training data sets of video data obtained from a set of users to derive the BP of each user in the training data, compare the derived BP with ground truth values of the respective user’s BP measured previously using an alternative mechanism and regress the estimated vital signs against the corresponding ground truth signs. In another embodiment, the at least one neural network includes a first neural network and a second neural network, and the at least one processor may be configured to provide the PPG signal to the first neural network to obtain a first approximation of the ABP signal waveform and to provide the approximated ABP signal waveform to the second neural network to refine the signal waveform and obtain the final representation of the ABP signal waveform. In some embodiments, to receive video data of at least a portion of the individual’s face, the at least one input is configured to establish a video communications channel between the individual and a medical practitioner and capture the video data during the video communication. In such embodiments, the system may comprise an image recorder to receive the video data, establish the video communications channel and capture the video data. In such instances, the at least one output may be configured to output the determined value of the at least one physiological parameter to at least one of: a medical practitioner and / or the individual in real time during the video communication. The at least one output may be configured to output to the individual and / or a medical practitioner, a message relating to an action to be taken by the individual and / or medical practitioner based on the at least one medical conclusion output. In some instances, the at least one processor may be configured to store the determined one or more blood pressure values in association with an identifier uniquely identifying the individual. In another embodiment, the at least one processor may be configured to retrieve data relating to one or more individual-specific properties of the individual, analyse the determined one or more blood pressure values based on the retrieved data and a set of predefined rules to determine a medical conclusion regarding hypertension, and the output may be configured to output the determined medical conclusion. In such instances, the processor may be further configured to determine the one or more blood pressure values multiple times over a period of time to generate a historical set of blood pressure values, store the historical set of blood pressure values in association with the unique identifier. To analyse the determined one or more blood pressure values, the at least one processor may be further configured to retrieve the historical set of values for analysis in combination with the determined one or more blood pressure values based on the retrieved individual-specific data and the set of predefined rules to determine the medical conclusion. In some instances, the processor may be configured to transmit (via the at least one output) the determined one or more blood pressure values to a remote healthcare provider system for storage and / or ongoing monitoring of the individual. Optionally, the processor may be further configured to determine a value of one or more additional physiological parameters of the individual from the PPG signal, the one or more additional physiological parameters comprising at least one of: heart rate (HR); respiratory rate (RR); and heart rate variability (HRV). Within the scope of this application it is expressly intended that the various aspects, embodiments, examples or alternatives set out in the preceding paragraphs, in the claims and / or in the following description and drawings, and in particular the individual features thereof, may be taken independently or in any combination. That is, all embodiments and / or features of any embodiment can be combined in any way and / or combination, unless such features are incompatible. The applicant reserves the right to change any originally filed claim or to file any new claim accordingly, including the right to amend any originally filed claim to depend from and / or incorporate any feature of any other claim although not originally claimed in that manner. BRIEF DESCRIPTION OF THE DRAWINGS One or more embodiments or aspects will now be described, by way of example only, with reference to the accompanying drawings, in which: Figure 1 is a schematic block diagram showing an overview of the main components of a computing system architecture arranged to implement processes for determining and monitoring one or more physiological parameters of an individual according to aspects of the present disclosure; Figure 2 is a flow diagram providing an overview of the processing steps involved in the process implemented using the system of Figure 1; Figure 3 is a schematic block diagram illustrating further details of the user device and computer system used in the system of Figure 1; Figure 4 is a flow diagram illustrating further details of part of the process of Figure 2, involving the determination of the physiological parameters of the individual from captured video data; and Figures 5A and 5B are flow diagrams illustrating further details of part of the process of Figure 2. Where the figures laid out herein illustrate embodiments of the present disclosure, these should not be construed as limiting to the scope of the disclosure. Where appropriate, like reference numerals will be used in different figures to relate to the same structural features of the illustrated embodiments. DETAILED DESCRIPTION A high level computing architecture system 1 that may be used to implement aspects of the present disclosure, and an overview of how a method 100 for determining and monitoring one or more physiological (human) parameters of an individual may be implemented using such a computing system, will now be described with reference to Figures 1 and 2. The computing architecture system 1 (hereafter simply referred to as the ‘system T) comprises a (mobile) user device 2a associated with a user 2b, who in this case is the individual whose physiological parameters are to be measured using the system 1. The user device 2a is arranged to capture, in Step 105 (of Figure 2), a video of at least a portion of the individual (e.g., of some or all of their face). A communications network 4 (such as the internet) is also provided which is used for communications between the various system components. The system also comprises a first (remote) computing system or server 6. The user device 2a is configured to transmit the captured video data to the first computing system 6 via the communications network 4. Thereafter, the first computing system 6 is configured to process the received video data and to extract, in Step 110, data from regions of interest (ROIs) of the captured images. The first computing system 6 is configured to process and analyse, in Step 115, the extracted video data to determine values of one or more physiological parameters of the individual in question. The system 1 comprises one or more databases I data stores 8 (although only one is shown for illustrative ease), with which the first computing system 6 is in association or in communication. The first computing system 6, in combination with the database(s) 8, is configured to assess the determined values of the physiological parameter(s) and to determine based on the parameter values, in Step 120, a medical conclusion / result for the individual. This result is then output, in Step 125, to the individual, for example via transmission of the result back from the first computing system 6 to the user device 2a via the communications network 4. The system 1 further comprises one or more second (remote) computing systems 10 (which itself take the form of one or more networked computers and / or servers, in association with one or more databases I data stores). The second computing system 10 may be associated with a healthcare provider; for example, in the case where the system 1 is to be implemented in the United Kingdom, the healthcare provider may correspond to the National Health Service (NHS). The first computing system 6 is configured, in Step 125, to store the determined physiological parameter(s), in the database 8 for example; and / or to forward the determined physiological parameter(s) to the second computing system 10. The medical conclusion / result may also be stored and / or transmitted along with the corresponding determined physiological parameter(s). In view of their respective functionalities in the current example implementation, and for ease of reference and distinction between the two systems in the present disclosure, the first computing system 6 will also be interchangeably referred to hereinafter as the ‘parameter monitoring system 6’; and the second computing system 10 will be interchangeably referred to hereinafter as the ‘healthcare provider system 10’. Figure 3 shows additional details of the user device 2a, the parameter monitoring system 6 and its associated database 8, that are used to implement the above described processes. The user device 2a may correspond to any computer or computing device that has the appropriate image capture, processing, transmission and display functionality to enable the corresponding steps of the above-described method 100 to be implemented. For example, the user device 2a may take the form of a mobile phone, smart (computing) device, laptop computer, desktop computer, tablet etc., with which the user 2b can interact. The parameter monitoring system 6 may take the form of one or more computers and / or servers, in communication with one another as necessary, and having the required receiving, processing, analysis, storage and transmitting functionality to implement the corresponding steps of the abovedescribed method 100. For example, the parameter monitoring system 6 may take the form of a distributed (e.g., cloud based) computing system, such as the Amazon Web Services (AWS), which provides on-demand cloud computing platforms and APIs to users as required. In such cases, it will be appreciated that the communications network 4 and the parameter monitoring system 6 may be integrated with each other. Now considering each of these components in more detail, the user device 2a comprises a processor 12 and a storage module I data store / database 14; these components together provide the processing and storage functionality for the user device 2a to implement aspects of the disclosure. Additionally, the user device 2a comprises an image recorder 16 (e.g., one or more cameras) which can be used to provide the desired video (image) capture functionality to capture a video data (a series of images over time) of at least a portion of the face of the user 2b. The image recorder 16 may form part of the user device 2a itself: for example, in the form of one or more cameras integrated into the user device 2a, operated via an associated application that is installed on the user device 2a and run by the processor 12 and storage 14 in combination. Alternatively, the image recorder 16 may take the form of an external camera I recorder (such as a webcam) that is in operative communication with the user device 2a, for example, via a wireless connection to the user device 2a; this external camera may also be operated using an application installed in the storage 14 of the user device 2a and run by the processor 12 of the user device 2a. The user device 2a also comprises a display or user interface 18 (GUI), which the user 2b may use to interact with aspects of the user device 2a, for example to cause (e.g., provide approval for) the image recorder 16 to begin recording a video of their face and to receive feedback as to the results of the physiological parameter monitoring process. The user device 2a also comprises a parameter monitoring application 20 (illustrated in the figure as ‘DocMe’) that is stored in storage 14; the functionality of this application is implemented by the processor 12. The parameter monitoring application 20 is configured to manage, in combination with the rest of the components of the user device 2a, the implementation of aspects of the disclosure relating to determining and monitoring physiological parameters of the user 2b. For example, the parameter monitoring application 20 may be configured to facilitate I manage the interactions of the user 2b with the user device 2a (e.g., via the image recorder 16 and the display 18) to ensure that the necessary video data of the user 2b can be captured in an appropriate manner. More specifically, in one example, the user 2a may be able to interact with the parameter monitoring application 20 directly via the display 18 and, under instructions provided by the parameter monitoring application 20, operate the user device 2a and utilise the image recorder 16 to record the video data of the appropriate portion(s) of their face. The user 2b may additionally or alternatively use the display 18 to interact with a different application or set of applications in the user device 2a in order to establish a video call (for example, via well-known video-calling technologies such as WhatsApp, Zoom or FaceTime) using the image recorder 16; the parameter monitoring application 20 may then operate simultaneously (in the background) in order to manage the subsequent processing and transmission of the output video data stream / segment that is captured by the image recorder 16. The parameter monitoring application 20 may also carry out its functions in combination with one or more other applications (not shown) that would also be maintained in the storage 14 and implemented using the processor 12 of the user device 2a. The user device 2a also comprises a communications engine 22, which may be provided in combination and in communication with the parameter monitoring application 20. The communications engine 22 is configured to manage and facilitate the communication of data between the parameter monitoring application 20 and the components of the parameter monitoring system 6. For example, the communications engine 22 is configured to handle the communication of the video data (segment), captured by the image recorder 16, to the parameter monitoring computing system 6; and to subsequently receive the values of the physiological parameter(s) and / or the medical conclusion result that is returned by the parameter monitoring system 6 after the processing and analysis of the input video data is carried out. As noted previously in relation to Figures 1 and 2, the parameter monitoring system 6 is configured to provide physiological parameter determination, monitoring and analysis functionality, in combination with the associated database 8. To this end, the parameter monitoring system 6 comprises the following components which provide various aspects of this overall functionality: a facial detector 24; a feature extractor 26; a PPG (Photoplethysmography) signal calculator 28; a physiological parameter calculator 30; and a medical conclusion I scoring engine 32. A video preprocessor 33 is also provided which is configured to carry out any pre-processing that may be required to render the video data suitable for subsequent processing and analysis. It will be appreciated that although these components and their respective functionalities are illustrated separately in Figure 3 for ease of reference subsequently, depending on the nature of the computing system architecture used to implement the desired functionality, these components and their associated functionality may be combined with one another as appropriate; they may also be implemented using one or more processors (not shown) in association with one or more storage components (also not shown). Additionally, where the parameter monitoring system 6 is implemented as a distributed cloud-based networked system, these components (and their associated functionalities) may be implemented by different processing and storage resources in the distributed system. The parameter monitoring system 6 also comprises one or more APIs (application programming interfaces) 34 which are configured to receive the incoming video data from the (communications engine 32 of the) user device 2a, and to subsequently return the output data - e.g., calculated parameter value(s) and / or the medical conclusion I result I score - to the user device 2a. These APIs 34 can also be configured for communication of some or all of the processed data and analysis results to the healthcare provider system 10. In more detail, the facial detector 24 is configured to process the incoming video data and to detect the presence of the user’s face in the captured video; the facial detector is further configured to detect specific portions, or defined regions of interest (ROIs) of the detected face and to extract the video data corresponding to those ROIs for subsequent downstream processing. The feature extractor 26 is configured to process the received video data and to ascertain various characteristics of the individual present in the video that can be associated with demographic information (e.g., age, sex / gender, heritage / ethnic origin etc.) of the user 2b. The PPG signal calculator 28 is configured to process the video data extracted from the ROIs of the user’s 2b face, and to implement one or more signal processing algorithms, using one or more base models, to convert this video data into a PPG signal. As is known in the art, a PPG signal corresponds to the temporal variation in the amount of light that is passing through the skin of the user 2b (in those ROIs), and therefore correspondingly provides an indication of arterial pulsation of the user 2b over time. The physiological parameter calculator 30 is configured to process and analyse the generated PPG signal and, using one or more filtering and processing algorithms, to calculate one or more physiological parameters of the user 2b. The calculated parameters include (but are not limited to): (1) blood pressure (BP) and specifically the systolic and diastolic blood pressure (SBP and DBP) as well as the mean arterial pressure (MAP); (2) respiratory rate (RR); (3) heart rate (HR); and (4) heart rate variability (HRV). The parameter monitoring system 6 is configured to store the calculated physiological parameter(s) in the database 8, in a data entry 36 which is (uniquely) associated with the user 2b in question. The medical conclusion / scoring engine 32 is configured to utilise one or more of the calculated physiological parameters to ascertain an associated medical conclusion and / or score. The resulting medical conclusion / score that is to be output subsequently is determined using a set of rules 38, stored in the database 8, which have been defined in line with the requirements of the healthcare provider. These rules are necessarily highly regulated and based on clinical care guidelines that are in line with best practices adopted by the healthcare provider. Additionally, the medical conclusion I score may be determined taking into account historical data of the user 2b, for example, values of the physiological parameter(s) that have been calculated over a predefined period of time preceding the current instance of parameter determination and stored in the user’s data entry 36. The demographic data ascertained by the feature extractor 26 may also be taken into consideration when generating the medical conclusion, and may also be stored in the corresponding data entry 36 in the database 8. Further details of a method 200 of determining the physiological parameter(s) will now be described with reference to Figure 4, and on the basis of implementation using the computing system architecture illustrated in Figure 3 - namely, using the parameter monitoring system 6. It will be appreciated, however, that this method may be implemented differently; for example, some or all of the components and their associated functionality described above as being implemented using the parameter monitoring system 6 may be implemented instead by the user device 2a itself (provided this device is provided with sufficient computer processing power and storage capacity). The method 200 defines a processing workflow that takes, as its input, video data captured of the user 2b I individual; and produces, as its output, values of one or more physiological parameters of interest of that individual. Prior to entry into the processing workflow (or as an initial pre-processing step), the incoming video data, captured using the image recorder 16, may be analysed and processed (for example, by the video pre-processor 33) to ensure that the video data that is to be subsequently used to calculate the physiological parameters is of sufficient quality and length for the processing to be carried out. The method 200 begins in Step 205 where the pre-processed video data (comprising a segment of video data of sufficient quality that is around at least 15 seconds in length, for example), enters the processing workflow. In initial processing Step 210, the received video data segment is analysed by the facial detector 24 and the feature extractor 26 in the corresponding manners described above. Specifically, the facial detector 24 implements a facial recognition algorithm to first identify and detect the presence of the user’s 2b face within the video data; and then subsequently, on a smaller scale, to pinpoint and define regions of interest (ROIs) of the user’s face, and to extract the video data specifically from those ROIs. This facial recognition algorithm may take the form of one of a number of currently implemented ‘off-the-shelf’ or open-source facial recognition algorithms I programmes that have already been trained to perform facial recognition on image / video data (for example the Google® MediaPipe machine learning solutions). It is useful to extract data from these ROIs of the user’s face, as it reduces the volume of data that needs to be processed thereafter to determine the physiological parameter(s), relative to if the entire image was to be processed; these ROIs also typically correspond to areas of the user’s face 2b where the colour changes in the skin tone can be more easily detected (e.g., where the skin is thinner and / or the blood vessels are closer to the surface of the skin, such as in the cheeks of the user 2b, as this allows the changes in light properties resulting from blood flow to be more readily monitored). Additionally (before, after or substantially simultaneously with the facial extraction processing), the feature extractor 26 analyses the received video data and implements a corresponding demographic detection and extraction algorithm (which may also take the form of one of a number of currently implemented ‘off-the-shelf’ or open source programmes) to determine various characteristics and features of the user 2b based on detectable properties I markers in their captured video data. For example, the age, sex I gender, ethnic origin group of the user 2b can be determined from the video data using the appropriate algorithms. Once the video data of the ROIs of the user’s face have been extracted from the initial (pre-processed video), the extracted data is processed by the PPG signal calculator 28 in order to extract the corresponding PPG signal. This processing begins in Step 215 where the colour space (red, green blue, or RGB) channels which make up the video data signals in those ROIs are isolated I extracted. The signals obtained in each of the RGB channels are then processed in Step 220 to obtain the PPG signal, using a base model that maps the RBG signals obtained from the video data of the user’s face to a corresponding PPG signal. Initially, the RGB channel signals are each standardised I normalised and filtered as appropriate. For example, in one illustrative embodiment, it is envisaged that the signal in each of the RGB channels is scaled such that the signal waveform oscillates around 0 and such that the standard deviation of the signal in that channel is set to around 1. The resulting signals are then each passed through a bandpass filter having frequency ranges that correspond to the expected resting heart rate and respiratory rates of an individual (for a healthy / normally-functioning human body). For example, in an illustrative embodiment, it is envisaged that a Butterworth bandpass filter may be applied to the signal in each of the RGB channels, with frequency rates of interest between around 0.75 and 3.0 Hz (corresponding to a resting heart rate for a functioning human body of between 45bpm and 180pm); and more particularly the range can also be limited further to between 0.75Hz and 1.7Hz (to focus on the lower frequencies and the corresponding lower resting heart rates). The filter may also be used to obtain a further set of frequency rates of interest - e.g., around 0.18 and 0.5 Hz - which correspond to an expected respiratory rate for a functioning human body. Thereafter, a PPG signal is recovered from each of the RGB channels via blind source separation using an independent component analysis (ICA) algorithm (e.g., the publicly available FastICA algorithm) which is well known to be suitable for separating out one or more specific wanted signals from an initial, input signal comprising both the wanted signal(s) and signals originating from a variety of noise sources. The resulting PPG signals recovered from each of the filtered RGB channels are then benchmarked and the best signal is selected to proceed with the subsequent downstream processing flow. Various different factors I conditions I properties can affect the quality of the resulting signal: for example, the type of light that is used to illuminate the individual; the pigment of the individual’s sign; the conditions of the surrounding environment (e.g., flickering of the illuminating light); and even the nature of the internet connection that is used to obtain the resulting video. The benchmarking process will take these factors into account and can select the most appropriate signal for further analysis (e.g., the one having the best signal-to-noise ratio and / or the most prominent features corresponding to the wanted signal). The processing workflow then proceeds to determine the values of one or more physiological parameters of interest from the PPG signals. The specific mechanism used to calculate the parameter(s) differs depending on the nature of the parameter itself. To begin with, a frequency analysis is performed in Step 225 on the PPG signal - where the PPG signal is mapped from the time domain into the frequency domain (for example, using a Fourier Transform or Fast Fourier Transform (FFT), or via use of a periodogram), so that frequency peaks in the signal can be identified. As mentioned previously, the PPG signal is a time-varying signal of the amount of light that is passing through the skin, and is hence a measure I indication of the arterial pulsation. The heart rate (HR), HR variability (HRV) and respiratory rate (RR) of the individual captured in the video, which are all linked to arterial pulsation, can then be calculated in Step 230 from the derived frequency signal. Specifically, since the arterial pulsation is driven by the beats of the heart, the primary arterial pulsation frequency will correspond to the heart rate (HR) of the individual. As a result, it will be appreciated that a dominant frequency component of the PPG signal will correspond to the heart rate; and in the frequency domain, this should then correspond to the component having the highest peak in the frequency range of interest for heart rates (i.e., the 0.75 and 3.0 Hz frequency range described above). Therefore, the HR value of the individual is determined by identifying the highest peak in the frequency domain of the PPG signal, and which lies within the above-mentioned frequency range. As the respiratory rate (RR) of the individual is linked to arterial pulsations and will also result in temporal variations in the amount of light passing through the skin, albeit at a different (slower) rate to the variations caused by the individual’s heart beats, a corresponding method to that set out above for the HR measurements can also be carried out in relation to RR determination. Specifically, the RR value of the individual is determined by identifying the highest peak in the frequency domain of the PPG signal, and which lies within the frequency range of interest for respiratory rates (i.e., the 0.18 and 0.5 Hz frequency range described above). As will be appreciated from the above description of heart rate, the frequencies of interest for measuring respiratory rate are significantly lower than those for measuring heart rate and hence the two peaks can be distinguished in the frequency doman. Finally, the HRV value of the individual can be calculated by determining the time intervals (e.g., in milliseconds) between successive heart beats of the individual over a predefined period of time -this can be determined by measuring the time intervals between successive peaks in the original time-domain PPG signal over that period of time. The HRV value is then calculated by taking the standard deviation of all of the calculated time intervals over that predefined period of time. In an illustrative example, the predefined period of time may be less than around a minute; more particularly, in some cases the predefined period of time may be around 40 seconds, and a similar amount of time may then be utilised to carry out the required processing. In some cases, the predefined period of time may be even less, and may even be as little as around 15 seconds, depending on the capabilities of the system carrying out the processing and the methods that are used. The processing workflow also includes a method for determining blood pressure (BP) values of the individual from the PPG signal that was derived in Step 220. As a starting point, it will be appreciated that there is a correlation between the speed of blood flow, and the resulting blood pressure, of a given individual. More specifically, when blood vessels are contracted, the blood within these vessels will flow more rapidly which will result in a higher measured blood pressure. Conversely, when the blood vessels are more relaxed, the blood flow within these vessels is slower and steadier which will result in a lower measured blood pressure. Studies have been conducted to investigate the rate of blood flow (also known in the art as Pulse Wave Velocity, or PWV) on this basis. A more detailed discussion of how the blood pressure can be derived from the PWV is provided in Annex A at the end of this description. To summarise, an Arterial Blood Pressure (ABP) signal waveform can be predicted from the PPG signal (due to the relationship between the blood pressure and the speed of blood flow, set out in the Annex) and values of interest of the blood pressure (e.g., minimum, maximum, mean) can be extracted from the resulting ABP signal waveform. Building upon this knowledge of the underlying theory, the method implemented by the parameter monitoring system 6 for determining the BP values of the individual from the PPG signal proceeds in the following manner. In Step 235, in order to improve the accuracy with which the subsequent ABP signal waveform can be derived, the input PPG signal can undergo pre-processing to filter out and / or attenuate various irregularities and noise sources in the PPG signal. This pre-processing may take the form of various well-known techniques, including but not limited to applying wavelet transforms to several decomposition levels; wavelet denoising; reconstruction of the denoised wavelet decompositions; mean normalization of the resulting signals. Thereafter, the PPG signal is used in Step 240 as an input to one or more ‘ABP signal estimate (derivation) networks’ in order to derive an estimate of the ABP signal waveform. In an illustrative embodiment, the ABP signal estimate (derivation) network(s) can be implemented as one or more convolutional neural networks (CNN), based on deep learning models and programmed with algorithms forming part of a ‘base model’ defining the relationship between the PPG signal and the ABP signal (e.g., as set out in Annex A). Using this network(s), the shape (e.g., amplitude, period etc.,) of the resulting ABP waveform can be approximated to the actual ABP of the individual. In some cases, the implementation may involve defining multiple models and correspondingly may involve the use of multiple ‘estimate networks’. For example, a preliminary (rough I first) estimate of the ABP signal waveform may first be derived using a first ‘Approximation Network’ (based on one model); this preliminary estimate may then be refined via further processing and the use of a second ‘Refinement Network’ (another convolutional neural network, or CNN, programmed with algorithms further defining I refining the relationship between PPG and ABP signal waveforms). Using this ‘Refinement Network’, the shape (e.g., amplitude, period etc.,) of the resulting ABP waveform is thereby refined to more closely approximate the actual ABP of the individual. However, in other cases, the use of multiple networks may not be necessary; a single model (and a single network) may be enough to obtain a sufficiently accurate approximation of the individual’s ABP from the PPG signal to allow the necessary analysis for determination of the key BP parameters. Once the final ABP signal waveform has been approximated I generated using the base model(s) and algorithms, analysis of this resultant waveform can be carried out in order to determine key BP parameters, such as the systolic blood pressure (SBP); diastolic blood pressure (DBP) and mean arterial pressure (MAP). As noted in Annex A, the SBP and DBP correspond to the maximum and minimum of the ABP signal; and the MAP corresponds to the mean value of the ABP signal. Considering the two CNNs described above in more detail, it is noted that these CNNs, and the models and algorithms that are provided to and used by these CNNs, to model, regress and derive the ABP signal waveform from the input PPG signal would have been initially trained using multiple training data sets of video segments obtained from various test subjects. In one example training implementation, 1 minute video segments capturing ROIs of various test subjects’ faces were captured over a period of time; simultaneously, corresponding ground truth values of the main physiological parameters that are to be determined in the present disclosure (HR, RR, HRV, and BP) were measured for these test subjects over the same period of time. These ground truth values were recorded using more ‘traditional’ mechanisms, such as by using cuff-based electronic sphygmomanometers and pulse oximeters. In order to increase the number of training and test data sets, individual training / test data pieces (also referred to as ‘episodes’) comprising a 10 second extract of any given captured video segment were produced. In an illustrative example of the training and testing methods that were adopted, a total of around 130,000 such episodes (comprising over 350 hours of video recorded time) were generated and utilised, with the test subjects spanning a range of ages between mid-forties into their eighties, and with a relatively even balance in gender between male and female. For training purposes, the CNN ‘ABP signal (derivation) estimate network(s)’ implementing an initial basic, best-guess algorithm and model was / were applied to obtain an approximate ABP waveform from an input PPG signal that was derived from each training data set; this approximate ABP waveform was then analysed and used to obtain an initial estimate of the corresponding vital signs. The accuracy of the waveform derived therefrom was then improved by regressing the estimated vital signs against the corresponding ground truth signs (by decreasing the value of the loss function). A corresponding process was implemented for testing purposes, where the trained CNN(s) was / were applied to the test data sets and the accuracy with which the waveform was derived was improved by regressing the estimated vital signs against the corresponding ground truth signs (by decreasing the value of the loss function). As a final step in the above-described physiological parameter determination method 200, some or all of the parameter(s)-HR, RR, HRV and BP-derived in Steps 230 and 245 above are then stored in Step 250 in the database 8, specifically in the data entry 36 corresponding to that user 2b. The appropriate data entry in which to store the determined parameters may be ascertained based on an identifier that is received from the user device 2a, along with the video data that is to be processed, and which uniquely identifies the user’s specific data entry. This identifier may take the form of an account or user ID associated with that user (e.g., an email address of the user 2b), and which has been set up by the user 2b, the mobile device 2a and / or the parameter monitoring system 6. A timestamp may also be added in association with the determined parameters when storing them in the database 8 for subsequent reference. Additionally, as part of this step, some or all of the determined physiological parameters may also be transmitted to the healthcare provider system 10 for monitoring, assessment and / or storage, with the option for retrieval at a later date (for example, for access by a healthcare practitioner for consultation or ongoing monitoring purposes). Additionally, or alternatively, some or all of the physiological parameters may be transmitted back to the user device 2a and stored / output by the parameter monitoring application 20. A method 300 for monitoring and providing feedback to an individual regarding their physiological parameters, and in particular providing feedback in relation to the BP values of the individual (determined using the above-described method 200), will now be described with reference to Figures 5A and 5B. This method combines the above-described methods, models and algorithms, implemented by the user device 2a and the parameter monitoring system 6 to calculate the individual’s BP values, with the use of highly regulated medical processing workflows that have been developed (based on clinical care guidelines) by healthcare provider systems. The method 300 begins in Step 305 with the receipt of incoming video data of the user device 2a; and proceeds in Step 310 with the processing of the video data (using the corresponding steps of the method 200 as described above in relation to Figure 4) to determine the BP values of the individual in question. Once the BP values have been determined, demographic data and other individual-specific data of the individual in question is ascertained in Step 315. For example, this may be ascertained using the feature extraction module 26 and as described above in relation to Figure 4. Thereafter, the database 8 can be accessed in Step 320 to retrieve stored historical data, for example physiological parameters - BP values - for that individual, which will be used to provide context to the newly ascertained BP values. If the individual is already known (for example, they already have an existing data entry 36 in the database 8, and this data entry contains previously-stored demographic data of the individual), this data can also simultaneously be accessed and retrieved from the database 8 to complement any demographic and / or individual-specific data that is ascertained in Step 315 from the recent input video data. More specifically, the historical BP values of the individual (spanning a predefined period of time, for example having been collected at intervals over the course of the last month) are retrieved and combined with the newly calculated BP values. The combined set of BP values (both historical and current) is then assessed in Step 325 in relation to the set of processing rules 38 stored in the database 8. These rules 38 have previously been generated by (or provided to) the system 6 in line with any specific regulatory requirements set and approved by the appropriate healthcare provider; they provide the framework for the processing workflow that can then be implemented in order to generate medical conclusions and advice for provision to the individual. Some examples of the advice that may be provided to the individual under various circumstances, which reflects the requirements of those rules, are set out in Table 1 below. - care *o * k >of sei' e.eA SnSs* Ar :^-^0-2155--½^^«.' Assto 2 -’ $5^0^0 \<x. o< t oc 3 w c_^J5±^’v ' t % '"■ "st r ' s v« o i « s cal 1 arct <f 'nmt Sae* lAdwoesiR?: in w stamen Ai Eintvbnv1-' (ACEo •>: '5* K CC5 rotors rec- ? a52e'.->2:tyr\>! neats:t rayc T^c 1 £. abates scC speac she?* scso* concerns eho^r h-jb tresi-yrt. o::ung oK-nr'karion: 'v?d'Canor'5 optics;:. Fa;;a: ".ookstwi yr.: Bissel pressure- sr if co-u hsvehaa anaavs-rr-e-l-Mg raacyon Vso canre-ac* more 3&xr hiss imsoo pressure B-re. P-asse gorovour S5 start an a£E ifskik'tox an recess ifye r&lsseratrce to A£tt. or :ACE; - i-sipopri, P±;r.fpm eaps-Bie: Table 1 - Example output messages following BP determination It should be appreciated that such advice is merely one example of how the processing workflow can be implemented in practice, and is not intended to be limiting to the scope of possible applications for this processing workflow. Furthermore, it will be appreciated that since the details of the workflow as a whole are defined in accordance with the requirements of a respective healthcare provider (and may be tailored I altered based on the region or country in which any given provider is based), the specific rules 38 that are applied when implementing the processing workflow may need to be altered accordingly to take into account the location of the individual. As such, one step in the workflow can involve ascertaining the geographical location of the user 2b (for example, using the IP address of the user’s device 2a) and using this knowledge to retrieve a specific set of rules as appropriate for subsequent use. An example implementation of a medical conclusion processing workflow that could be followed according to an illustrative example begins in Step 330 where it is ascertained whether the combined set of BP values comprises a maximum value that is within either a ‘Healthy’ or ‘High Risk’ range of BP values, as defined in the rules 38. For example, according to various established medical studies that identify and define thresholds for ‘healthy’ or ‘at-risk’ BP values, a ‘Healthy’ set of values corresponds to BP values where the maximum SBP value is less than 140 and the maximum DBP is less than 90; by comparison an ‘At Risk’ set of values corresponds to BP values where the maximum SBP value is greater than or equal to 180 and the maximum DBP is greater than or equal to 110. If it is determined that the combined set of BP values comprises a maximum value that is within either one of the ‘Healthy’ or ‘High Risk’ range of BP values, the processing workflow continues to Step 335, in which either a ‘Healthy’ or ‘Emergency’ outcome message respectively (with corresponding wording along the lines set out in Table 1 above) is scheduled to be provided to the individual as appropriate. The appropriate message and feedback is then provided to the user 2b. In some cases, this may be implemented by retrieving I generating the appropriate message and transmitting it to the user device 2a for provision via the display 18; where the network communication and processing capabilities of the system permit, this generating and display may be carried out in substantially real-time - i.e., during the course of a video communication with the user 2b (whilst ongoing video data is being recorded, or during the course of a video call that the user 2b is having with their healthcare provider and during which the video to be processed is captured). Alternatively, the message may instead be transmitted to a third party (for example a healthcare practitioner) and then presented to the individual thereafter. Subsequently, the newly calculated BP value is stored in Step 340 in the data entry 36 along with the rest of the historical data (along with a timestamp if appropriate to indicate the age of the stored values). It will be appreciated that the historical data can provide context to the BP values that are determined in the most recent set analysis; and the historical data can also be used to ascertain I highlight any patterns or trends that may be associated with the BP values over time. For example, if the most recent set of BP values is relatively high, but there is a trend of higher BP values across the historical data, this may be used to alter or provide context to the outcome that is recommended. This alteration and contextualisation may be carried out as part of the processing workflow itself; or it may simply be used for reference by the individual themselves (or by any third parties such as medical personnel) to alter the conclusions I recommendations that are provided. If the combined set of BP values does not have a maximum value that falls into either of these two ranges, the workflow then proceeds to Step 345 (in Figure 5B), where the combined set of BP values is reviewed in view of the demographic and / or individual-specific data. This additional data is used to provide further context to the combined set of BP values, and to aid in the decisioning regarding the recommendation I advice that is to be provided to the individual subsequently. Specifically, in an illustrative embodiment, the demographic data can be used in Step 350 to ascertain if the individual is associated with any risk factors in relation to certain BP medications that may be prescribed. Depending on the outcome of this demographic data context double-check, the processing workflow continues to Step 355 in which either a ‘Medical T or a ‘Medical 2’ outcome message (with wording along the lines set out in Table 1 above) is scheduled to be provided to the individual as appropriate. For example, if it is determined from the demographic data that the individual is above a certain age (e.g., above 55), one message outcome (specifically, the ‘Medical T message) is scheduled for provision. Alternatively, if the individual is not above a certain age but is of a particular ethnic origin (e.g., Afro-Caribbean origin), a different message outcome (specifically, the ‘Medical 2’ message) is scheduled for provision instead. This is because certain demographics (e.g., age, sex, ethnicity) have been ascertained, from previous medical research, to have a significant effect on and correlation with blood pressure of the individual, as well as in relation to the treatment that can I should be provided to the individual. As was the case for the ‘Healthy’ and ‘High Risk’ BP outcomes, the message outcomes are provided to the user via corresponding mechanisms to those described above. Finally, the newly calculated BP value is stored in Step 360 in the data entry 36 along with the rest of the historical data (along with a timestamp if appropriate to indicate the age of the stored values). Where the calculated BP values can be clearly seen to fall well outside the acceptable range of values for a functioning human body, this may indicate errors during the taking of the measurement (and / or an incompatible environment). When this occurs, the user 2b will be prompted to retake the measurements until an acceptable value is obtained; or until a predetermined threshold number of unacceptable values is obtained at which point the processing is terminated and the user 2b is advised to seek assistance. It will be appreciated that a result of implementing the above-described method 200 is the creation of what is essentially a ‘vital signs trend library’ for each individual. Where the data in the library is then subsequently monitored, for example when using historical data and contextual data based on the above-described method 300, in order to assess the progress of a (suspected) chronic disease, this would allow for multiple beneficial outcomes to be achieved. These include but are not limited to: remote identification of acute exacerbations of an individual’s chronic condition (e.g. Pneumonia in COPD); allowing contextual understanding of changes in disease progress; ensure general health monitoring; and ensure any adverse drug reactions, such as an allergy or an interaction, would be identified as early as possible. The benefits obtainable via the digital monitoring achieved by implementing the methods and systems described herein could improve confidence in remote prescribing and better care decisions. Multiple benefits in healthcare applications can be realised via the development and implementation of a clinical advice platform integrated with vital signs, as in the case of the processing workflow set out above. For example, the rising number of chronic diseases and ageing populations, multimorbidity and patients with complex needs, can be dealt with more easily via implementation of such a platform and workflow. Further benefits can be achieved in relation to decreasing the cost whilst implementing quality measures to improve patient outcomes, reducing variations in clinical care, and optimising the time of a given practitioner and maintaining a fluid workforce (e.g. optimising rotation patterns, reporting on targets and demonstrating compliance). Use of the above processing workflow provides contextual interpretation of vital signs data that is consistent with principles developed in ambulatory blood pressure monitoring, where concerns about the possibility of pathological blood pressure readings are evaluated in the context of a longitudinal data set. As a result, the above-described methods and systems are envisaged to be useful for managing chronic hypertension directly, monitoring blood pressure as a chronic disease risk factor and informing prescription pill changes. Some medications require 'normal' performance in some vitals, such as Blood Pressure, and hence the ability provided by the methods and systems described herein to monitor the various vital signs could also improve the provision of remote monitoring and prescription management. The combination of (clinical) guidelines, vital sign data and electronic health records (e.g., the above-described systems and methods implementing the clinical processing workflow method 300) will enhance the holistic support and patient-centred care, and will also make sure support available to telehealth (remote) clinicians. Further detailed description of one non-limiting example implementation for the above-described clinical I medical conclusion processing workflow, and the specific recommendations that may be provided thereby, is set out in Annex B. This description focuses on the specific example of how the above-described processing workflow could be integrated with the BMJ Best Practice clinical decision support tool (https: / / bestpractice.bmi.com / infoA, and specifically in relation to the ongoing monitoring of hypertension in an individual; however, it will be appreciated that other platforms I tools (especially those used at Point-of-Care), as well as other medical conditions of individuals, could also potentially benefit from implementation of the above-described processing workflow in a corresponding manner. For background, BMJ Best Practice is a clinical decision support tool that has been selected by NICE to provide ‘actionable knowledge’ to support decision making at the point of care e.g. at the bedside, in the ward, in the clinic, and in community settings, including the patient’s or carer’s home. It is uniquely structured around the patient consultation with advice on symptom evaluation, test ordering and treatment approach. The determination of the vital signs that has already been carried out by the methods and systems described herein can be used to help inform the practitioner as to how best to deploy the BMJ Best Practice guidelines for optimising blood pressure management. Many modifications may be made to the above examples without departing from the scope of the present disclosure as defined in the accompanying claims. For example, it is noted that portions of the user’s face are not the only locations on the user’s body that can be used as ROIs for data extraction and vital sign measurement using the methods described above. In fact, provided sufficient light can pass through the individual’s skin, and there is sufficient blood flow under the skin, other parts of the individual’s body can be used as ROIs for some or all of the vital sign measurements. In particular, it is envisaged that the fingertips and / or palms of the individual could potentially be used as ROIs, either in addition (or as an alternative) to the individual’s cheeks. As noted above, the user (subject) interface mechanism for these clinical processing workflows corresponds to the parameter monitoring application 20. All of the above-described processing (of methods 200 and 300) are carried out in the background by the components of the parameter monitoring system 6 and the results are provided to the user via the application 20 (and the display 18) if necessary. The integration between (1) the collection of video data and subsequent presentation of information I results to the user 2b, and (2) the underlying processing flows that are carried out using the clinical guidelines in order to determine and analyse the user’s physiological parameters, is facilitated by the provision of the communications engine 22 and the API 34. Flexibility is also provided thereby in how the physiological parameters can be stored in the database 8 and / or provided to and displayed I stored by the healthcare provider system 10, for example in an Electronic Health Record (EHR) system. It will also be appreciated that where the storage and / or processing capabilities of the user device 2a are appropriate, some or all of the processing, storage and / or output described above may be implemented in the user device 2a. Overall, implementation of the systems and methods described herein - integration of best practice clinical guidelines into a clinical processing workflow, informed by the physiological parameters derived as set out herein - provides multiple benefits. These include, but are not limited to: reduced time within any given consultation (both in person and video); elimination of unreliable vital signs readings; screening for unsuspected respiratory and cardiovascular abnormalities; remote triage of patients requesting home visits (for example, via integration of the functionality provided into the NHS 111 service in the UK); and reduction of patients attending A&E when unable to access GP appointments. Moreover, it is envisaged that the above-described integration of health care professional communication, with automation of repetitive GP tasks such as sick notes and prescriptions, can offer efficiency saving to optimise the use of the time the GP does have with the patient. ANNEX A - DERIVATION OF BLOOD PRESSURE VALUES FROM BLOOD FLOW RATE Studies have been conducted to investigate the rate of blood flow, which is popularly termed Pulse Wave Velocity (PWV). Their correlation can be represented as: E.h PWV = --- ^■r.p where E is the elastic modulus of the arterial wall, h is the thickness of the artery, r is radius of the artery and p is the density of blood in the artery. Based on PWV, blood pulses require a time delay to reach the periphery of the body from the heart, which is denoted as Pulse Transit Time (PTT). The relationship between PWV and PTT can be represented as: d PWV = -- PTT where PTT is the interval between a pulse wave being detected by two sensors and d is the distance between the sensors and the artery. In the first equation above, the elastic modulus E was assumed to be a constant when in fact the value of E in the artery is testified to be exponentially escalated with the blood pressure, as follows: E(P) = EoeaP where P denotes the blood pressure, Eo denotes the elastic modulus at 0 mmHg (the unit of blood pressure) and a is a parameter larger than zero that is closely related to arterial stiffness. The stiffer the artery, the greater the value larger than zero that is closely related to arterial stiffness. We can find a nonlinear relationship between blood pressure P and PTT, after we substitute the first and second equations into the third equation, to be: / 2\ / 1\ 2.r.p ~ p = - - ,ln(PTT)+ - ,ln(---7-.D2) \a) \a) Eoh In the method described in the rest of the document, a waveshape of the continuous Arterial Blood Pressure can be predicted from the PPG signal. After constructing the waveform of the blood pressure signal, the values of interest, namely systolic blood pressure (SBP), diastolic blood pressure (DBP) and mean arterial pressure (MAP) can be computed as follows: SBP = max(ABP); DBP = min(ABP); MAP = (ABP) ANNEX B - DETAILED DISCUSSION OF EXAMPLE USE CASE FOR MONITORING OF SUSPECTED HYPERTENSION BMJ Best Practice is funded by Health Education England, and freely available on web and app to all NHS staff. Since April 2019, BMJ Best Practice has been working in partnership with Health Education England to raise awareness and use of BMJ Best Practice nationally. Although the campaign has had success there is still work to do. BMJ Best Practice is a generalist point of care tool particularly useful for junior doctors, multidisciplinary teams, specialists working outside of their specialty and GPs. It is uniquely structured around the patient consultation with advice on symptom evaluation, test ordering and treatment approach. The BMJ BP search widget allows users to search BMJ Best Practice topics and clinical information directly from the EHR system. This gives fast access to trusted information from within the clinical workflow. The BMJ BP ‘HL7 info button’ provides links to specific BMJ Best Practice topic pages from patients’ problems list in the Electronic Patient Record (EPR). Essential hypertension is typically diagnosed by screening of an asymptomatic individual. Treatment of uncontrolled hypertension reduces the risks of mortality and of cardiac, vascular, renal, and cerebrovascular complications. Lifestyle changes are recommended for all patients: weight loss, exercise, decreased sodium intake, Dietary Approaches to Stop Hypertension (DASH) diet, and moderation of alcohol consumption. Choice of drug therapy is often driven by considerations related to comorbid disease, but achievement of blood pressure goal may be accomplished with a variety of therapeutic agent(s). Further details of the decisioning that can be carried out with reference to the BMJ Best Practice information are now described, with reference to the blood pressure targets set out in the table below: Clinic BP                ABPM / HBPM Age <80 years 140 / 90 mmHg 135 / 85 mmHg Age >80 years : 150 / 90 mmHg : 145 / 85 mmHg Managing hypertension recommendations: • A low salt diet is recommended, aiming for less than 6g / day, ideally 3g / day. The average adult in the UK consumes around 8-12g / day of salt. A recent BMJ paper showed that lowering salt intake can have a significant effect on blood pressure. For example, reducing salt intake by 6g / day can lower systolic blood pressure by lOmmHg • Caffeine intake should be reduced • The other general bits of advice remain: stop smoking, drink less alcohol, eat a balanced diet rich in fruit and vegetables, exercise more, lose weight If it is determined that ABPM / HBPM >= 135 / 85 mmHg (i.e. stage 1 hypertension), the following decisioning is carried out: • treat if <80 years of age AND any of the following apply; target organ damage, established cardiovascular disease, renal disease, diabetes or a 10-year cardiovascular risk equivalent to 10% or greater • in 2019, NICE made a further recommendation, suggesting that we should 'consider antihypertensive drug treatment in addition to lifestyle advice for adults aged under 60 with stage 1 hypertension and an estimated 10-year risk below 10%.This seems to be due to evidence that QRISK may underestimate the lifetime probability of developing cardiovascular disease If it is determined that ABPM / HBPM >= 150 / 95 mmHg (i.e. stage 2 hypertension), the following decisioning is carried out: • offer drug treatment regardless of age (taking into account flow chart below) For patients <40 years consider specialist referral to exclude secondary causes. Flow chart showing the management of hypertension as per current NICE guidelines Step 1 treatment • patients <55-years-old or a background of type 2 diabetes mellitus: ACE inhibitor or a Angiotensin receptor blocker (ACE-i or ARB): (A) o angiotensin receptor blockers should be used where ACE inhibitors are not tolerated (e.g. due to a cough) • patients >= 55-years-old or of black African or African-Caribbean origin: Calcium channel blocker (C) o ACE inhibitors have reduced efficacy in patients of black African or African-Caribbean origin are therefore not used first-line Step 2 treatment • if already taking an ACE-i or ARB add a Calcium channel blocker or a thiazide-like Diuretic • if already taking a Calcium channel blocker add an ACE-i or ARB or a thiazide-like Diuretic o for patients of black African or African-Caribbean origin taking a calcium channel blocker for hypertension, if they require a second agent consider an angiotensin receptor blocker in preference to an ACE inhibitor • (A + C) or (A + D) or (C + A) or (C + D) Step 3 treatment • add a third drug to make, i.e.: o if already taking an (A + C) then add a D o if already (A + D) then add a C • (A + C + D) Step 4 treatment • NICE define step 4 as resistant hypertension and suggest either adding a 4th drug (as below) or seeking specialist advice • first, check for: o confirm elevated clinic BP with ABPM or HBPM o assess for postural hypotension. o discuss adherence • if potassium <4.5 mmol / l add low-dose spironolactone • if potassium >4.5 mmol / l add an alpha- or beta-blocker Finally, patients who fail to respond to step 4 measures should be referred to a specialist.

Claims

1. A computer implemented method for measuring blood pressure values of an individual, the method comprising:receiving video data of one or more regions of interest ROIs of the individual’s face;processing at least one colour channel of the captured video data to identify and obtain a time-varying photoplethysmography PPG signal;providing the PPG signal as an input to at least one neural network to derive a representation of a signal waveform corresponding to the arterial blood pressure ABP of the individual;determining one or more blood pressure values of the individual from the ABP signal waveform; andoutputting the determined one or more blood pressure values to enable the blood pressure of the individual to be monitored.

2. The method of claim 1, wherein determining one or more blood pressure values comprises at least one of: (i) determining a systolic blood pressure SBP corresponding to a maximum of the ABP signal waveform; (ii) determining a diastolic blood pressure DBP corresponding to a minimum of the ABP signal waveform; or (iii) determining a mean arterial pressure MAP corresponding to an average of the ABP signal waveform.

3. The method of claim 1 or claim 2, wherein processing at least one colour channel of the captured video data comprises:processing a plurality of colour channels comprising at least red, blue and green colour channels;filtering the PPG signal in each of the plurality of colour channels to attenuate irregularities and noise sources; andbenchmarking the filtered PPG signal in each of the plurality of colour channels to select the best channel for subsequent analysis.

4. The method of any preceding claim, wherein the at least one neural network corresponds to one or more convolutional neural networks CNNs that have previously been trained by machine learning to derive the ABP signal waveform from the input PPG signal.

5. The method of claim 4, further comprising a prior process of training the one or more CNNs, the training process comprising:programming the one or more CNNs with an algorithm defining an initial relationship between the PPG signal and the ABP signal waveform; andrefining the algorithm of the one or more CNNs by:applying the one or more CNNs to training data sets of video data obtained from a set of users to derive the BP of each user in the training data;comparing the derived BP with ground truth values of the respective user’s BP measured previously using an alternative mechanism; andregressing the estimated vital signs against the corresponding ground truth signs.

6. The method of any preceding claim, wherein the at least one neural network comprises a first neural network and a second neural network, and wherein providing the PPG signal to the at least one neural network comprises:providing the PPG signal to the first neural network to obtain a first approximation of the ABP signal waveform; andproviding the approximated ABP signal waveform to the second neural network to refine the signal waveform and obtain the final representation of the ABP signal waveform.

7. The method of any preceding claim, wherein receiving video data of at least a portion of the individual’s face comprises:establishing a video communications channel between the individual and a medical practitioner; andcapturing the video data during the video communication.

8. The method of claim 7, wherein outputting the determined value of the at least one physiological parameter comprises:outputting the determined value of the at least one physiological parameter to at least one of: a medical practitioner and / or the individual in real time during the video communication.

9. The method of claim 8, wherein outputting the at least one medical conclusion comprises: outputting, to the individual and / or a medical practitioner, a message relating to an action to be taken by the individual and / or medical practitioner based on the at least one medical conclusion output.

10. The method of any preceding claim, wherein outputting the determined one or more blood pressure values further comprises:storing the determined one or more blood pressure values in association with an identifier uniquely identifying the individual.

11. The method of any preceding claim, further comprising:retrieving data relating to one or more individual-specific properties of the individual;analysing the determined one or more blood pressure values based on the retrieved data and a set of predefined rules to determine a medical conclusion regarding hypertension; andoutputting the determined medical conclusion.

12. The method of claim 11, further comprising:determining the one or more blood pressure values multiple times over a period of time to generate a historical set of blood pressure values,storing the historical set of blood pressure values in association with the unique identifier;and wherein:analysing the determined one or more blood pressure values comprises retrieving the historical set of values for analysis in combination with the determined one or more blood pressure values based on the retrieved individual-specific data and the set of predefined rules to determine the medical conclusion.

13. The method of claim 11 or claim 12, wherein the retrieved individual-specific data comprises one or more of the following: an age of the individual; a gender of the individual; a weight of the individual; a demographic profile of the individual; or a socio-economic profile of the individual.

14. The method of any of claims 11 to 13, wherein the predefined set of rules is based on clinically approved assessment guidelines, optionally comprising at least one of: BMJ Best Practice guidelines.

15. The method of any preceding claim, further comprising:transmitting the determined one or more blood pressure values to a remote healthcare provider system for storage and / or ongoing monitoring of the individual.

16. The method of any preceding claim, further comprising determining a value of one or more additional physiological parameters of the individual from the PPG signal, the one or more additional physiological parameters comprising at least one of: heart rate (HR); respiratory rate (RR); and heart rate variability (HRV).

17. A system for measuring blood pressure values of an individual, the system comprising:an input configured to receive video data of one or more regions of interest ROIs of the individual’s face;at least one processor configured to:process at least one colour channel of the captured video data to identify and obtain a time-varying photoplethysmography PPG signal;provide the PPG signal as an input to at least one neural network to derive a representation of a signal waveform corresponding to the arterial blood pressure ABP of the individual; anddetermine one or more blood pressure values of the individual from the ABP signal waveform; andand an output configured to output the determined one or more blood pressure values to enable the blood pressure of the individual to be monitored.33

Images