Cuff less estimation of blood pressure by PPG signals using machine learning

A novel method for the continual, cuff-less estimation of the systolic blood pressure (SBP) and diastolic blood pressure (DBP) values based on signal complexity analysis of the photoplethysmogram (PPG) is reported. The proposed framework estimates the blood pressure (BP) values obtained from signals

Project Title

Cuff less estimation of blood pressure by PPG signals using machine learning

Project Area of Specialization

Artificial Intelligence

Project Summary

A novel method for the continual, cuff-less estimation of the systolic blood pressure (SBP) and diastolic blood pressure (DBP) values based on signal complexity analysis of the photoplethysmogram (PPG) is reported. The proposed framework estimates the blood pressure (BP) values obtained from signals generated. Here in, the physiological signals were first pre-processed, then visualize the signal followed by the extraction of complexity features from PPG signals. Subsequently the complexity features were used in regression models Gaussian Process Regression (GPR) to predict the BP. The performance of the approach was evaluated by calculating the mean absolute error and the standard deviation of the predicted results. Complexity features from the PPG were investigated, along with the combined dataset. It was observed that the complexity features obtained from the combination PPG signals resulted to an improved estimation accuracy for the BP. Blood pressure is a main problem that has cost a lot of lives and there are methods to estimate it and observe and predict its working, but they are not as accurate as want them to be, so machine learning has to introduced in estimating such a disease. The photoplethysmography (PPG) has become widely recognized as a low-cost non-invasive detection technology for CVDs. The cardiovascular parameters detected using PPG technology include heart rate, blood oxygen saturation, blood pressure, assessment of arterial stiffness, and pulse wave velocity, among others. The PPG signal includes information on the hemodynamic process, hemorheology, and tissue status of the peripheral microcirculation system in the human body. That is, the PPG signal is an aggregated expression of many physiological processes in the cardiovascular circulation system. A physiological information database with high precision and a high sampling rate is urgently needed in PPG technology research in order to extract more cardiovascular parameters for the early screening and diagnosis of CVDs. We provide here a database containing physiological information and PPG waveform data collected over a year that can be used to research arterial blood vessel aging, arterial blood pressure detection, and screening of hypertensive and diabetic patients based on PPG signals.

Project Objectives

Now as we have mentioned the main problem that we are about to tackle, our main objective is to replace the traditional and old methods mainly the sphygmomanometer measurement method to estimate the blood pressure levels and that could be only done if we introduce the machine learning techniques in measuring it. PPG signal can effectively estimate blood pressure which has been recently studied improved by researchers. However, there are some limitations with collecting ECG and PPG simultaneously using a mobile phone. For these reasons, and for simplicity, a few researchers have attempted to estimate blood pressure based using only PPG signals. The concept of estimating blood pressure (BP) using only PPG signals seems to promising and is optimally implemented when the PPG signal is of high quality. Therefore, providing a database that can help with estimating BP using only PPG will help further research in this area. We need datasets in good quantity and we have to run those datasets on the Gaussian Process Regression model which is one of the best methods for estimating the BP and giving results accurately. According to the experiment, the accuracy of GPR predictions of systolic blood pressure values exceeded 90%. Our experimental results should be above 90%. So our main aim will be performing our project with this model. The dataset collection program involved acquiring information on the basic physiology of individuals, extracting information on cardiovascular diseases from hospital electronic medical records, collecting PPG waveform signals, and detecting instant arterial blood pressure at the same time.Now the project we are working on has no doubt a huge scope. Why is that? Let me explain. Blood pressure disease is a harmful one and can cause casualties and can lead to many other diseases which is a big problem, especially in Pakistan. As quite a lot of people in Pakistan die because of Hypertension. The National Health Survey of Pakistan estimated that hypertension affects 18% of adults and 33% of adults above 45 years old. In another report, it was shown that 18% of people in Pakistan suffer from hypertension with every third person over the age of 40 becoming increasingly vulnerable to a wide range of diseases.

Project Implementation Method

In this research paper (Blood Pressure Estimation Using Photoplethysmogram Signal and Its Morphological Features) there were a lot of features taken in order to read a signal and extract it but we after studying in detail about Machine learning and its feature detection, it is not necessary to maximize the features in examining a signal. So in our proposed method we used these features, and they are Age, Heart rate, weight, Height, BMI/t1, Weight/tpi, Weight/tpp, Weight/t1, and BMI/tpp. Now the main reason we used these features because in machine learning the results always shows accuracy when the features taken are correlated to each other, and in our method the above mentioned features are correlated and we have also found that with finding these features we get the best results, hence, it’s not necessary to find a bundle of features in order to read a signal. The main aim for accurate artificial intelligence predictions depends on correlation of the features and having the best model for these features which we are using in this project and that is Gaussian Process Regression (GBR).

Our project is mainly based on this research paper (Estimating Blood Pressure from the Photoplethysmogram Signal and Demographic Features Using Machine Learning Techniques). As we have taken our selected features from this paper and also got the idea how to extract these features by studying their works and methods. They also have taken tests and got the signals in which some were fit and some were unfit. The unfit PPG signals were ignored because they had some noise in them, so the fit PPG signals were studied and were used in preprocessing, then feature extraction, after that we create the CSV file and put these features in that CSV and then we do the process of checking the file if any data is missing, after this we clean the data and get the projected results.

Benefits of the Project

Patient outcomes have now taken the place from products and services as the main focus of healthcare providers. Medical organizations are pressing down on this trend and are now implementing advanced technologies like machine learning in healthcare to progress patient care and patient outcomes. As a result, the use of machine learning in healthcare is gradually but increasingly revolutionizing the healthcare industry. While the ability to record massive amounts of information about individual patients is transforming the healthcare industry, the volume of data being gathered is impossible for human beings to evaluate. Machine learning offers a way to automatically find patterns and examine unstructured data. This allows healthcare professionals to move to a personalized care system, which is known as precision medicine. Now the system we have proposed is basically more of an automated method of measuring and estimating blood pressure and observing its values, which changes every second. This automated method that we have proposed is doing estimation through Machine Learning. ML usage has increased in the medical sector as poor health is the most important issue in our lives. People have lost their lives due to various diseases throughout the course of our history and yet every single   time we ask ourselves about creating a medicine that can cure such a disease or create an electronic instrument or appliance to treat an individual or a group of people.

Machine learning is the best possible solution in helping us tackle this disease of hypertension, as many people are losing their lives on daily basis in Pakistan. Every third person in Pakistan is effected from BP disease as it is the root cause of many other diseases, which could take someone’s life very quickly and easily. In Machine learning here we can get great accuracies which can be a far better method of estimating the BP values than the current and old method of measuring BP through sphygmomanometer.  The machine learning model that we have chosen for our system is Gaussian Process Regression. Gaussian process regression (GPR) is a nonparametric, Bayesian approach to regression that is making waves in the area of machine learning. GPR has several benefits, working well on small datasets and having the ability to provide uncertainty measurements on the predictions.

Technical Details of Final Deliverable

Software:

• MATLAB
• EXCEL
• PYTHON

We got the PPG signals in raw form and the with the help of matlab we did its preprocessing and also did its feature extraction through matlab, so matlab was a requirement in our project in dealing with signals.

Then we required Excel software in which we could store the entries.

We also need python, which is one of the most important tools that is used for the analysis of the graph in great detail and making the work comparatively easier for us.

Hardware:

• Sensors for PPG
• Raspberry Pi

Then we also need sensors which would take PPG signals and that would be implemented on our raspberry pi.

Now the data we will get from the sensors will be put into the model, so that in return would give us our requirement of systolic and diastolic blood pressure values.

In our case, there will be a PPG sensor that will be connected onto the patient for 2.1 seconds. From that connection of the sensor we will get a PPG signal. That PPG signal would ultimately be used to measure the Blood Pressure values.

So in our proposed method there will be just 2.1 seconds of the patient needed to get there BP values. For continuous measurements we can merge our system into a smart watch or we can get the BP values on our cell phone. So in this case, it would make the job quite easier for both the patient and the doctor.

So in short, the main connection with the patient is of the sensor that would give us PPG signals.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Medical

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Good Health and Well-Being for People

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com