Communication between the Point of Care (PoC) system and the mobile devices happens over bluetooth and for all other communications, our cloud services are used. The machine learning tools along with the database are deployed in cloud. The patient can access the data from cloud through a web-based or mobile client. An easy to use android app and web client will be provided for this purpose.
The initial design of the device has been conceptualized after intense discussion within the team (medical experts at AIIMS, New Delhi, Technology experts at IIT Kharagpur). The screening device will be designed in the form of a chair equipped with various biosensors and health parameter measuring components. Unfortunately, UIs of the most of medical devices are generally complicated and ill-designed. Often following simple design principles can improve the user experience. The UI should be guided by patient's language with minimal usage of his/her cognitive load. The ergonomic design and overall appearance of the device resembles a comfortable chair that appears welcoming and somewhat enticing to the visiting patients, such that they would be willing to undergo a voluntary health screening.
The 5-lead ECG signal will be derived from a patient with/without the patch electrode at the sitting position. In order to enhance the patient’s comfort, a contact-free ECG can be designed. Here the patient will be sitting on a metal chair and the measurement will be conducted from the chair itself. The tradeoff for the enhanced patient’s comfort is poor Signal-to-Noise Ratio (SNR) for the contactless measurement.
Spirometry is the most common of the lung function tests. In this work, the respiratory (airflow) signals of both COPD patients and healthy individuals may be acquired. For the real measurement, a differential pressure based air-flow sensing mechanism can be employed. Differential pressure flowmeters introduce a constriction in the pipe that creates a pressure drop across the flowmeter. When the flow increases, more pressure drop is created. The sensor (shown in figure) will be placed on the two sides of a constriction and the pressure difference will be converted to the voltage output suitably calibrated with the airflow.
A Resistance Temperature Detector (RTD)-based body temperature sensing mechanism may be deployed with the user-friendly beep after the once the measurement is over. A bridge followed by opamp-based amplifier circuit is the usual candidate here.
In order to measure the weight of the user, a strain gauge-based load cell unit is proposed. Alongwith this the Body Mass Index will also be computed, if the height of the user is available. For height information, either it may be measured by low-cost LASER-based range finder or it may be taken as the input from the patient.
Electronic stethoscopes converts the acoustic sound wave obtained through the chest piece into electrical signals via microphone. This signal is then transmitted through amplifier and processed for optimal listening. A Littman Cardiac stethoscope is an available product on this line. Its integration with our proposed equipment is an important topic of research.
The estimation of SpO2 level (percentage of oxygenated haemoglobin compared to the total amount of haemoglobin in the blood ) is possible using the PPG signal acquired from the fingertip in a transmission mode. The measurement needs to be taken differentially (over two wavelengths) with pulsed sources. The challenge of the PPG is that the subject should stay still so that no moving artifact involves.
The cuffless blood pressure measurement will be performed (after suitable calibration) via ECG and PPG signal as per the IEEE Std 1708. The measurement scheme is usually based on Pulse Transit Time.
We are planning to analyze the multimodal data consisting of ECG, airflow (respiratory) data, body temperature, body weight, digital Stethoscope-based acoustic data, PPG (SpO2) data and blood pressure data.
The vast amount of data captured by the delivery system and stored in cloud storage is required for analysis of the data for disease monitoring and prevention. Two inter-related supervised binary classification frameworks need to be developed for (i) CVD classification and (ii) COPD classification.
Our Desease Progression Model (DPM) uses data analytics tools to time sequence medical data of individuals to develop a model for disease progression.
The efficacy of the solution significantly depends on the knowledge shared by the medical expert as an input for designing the computer algorithms to create analytical models for disease classification and progression. Validation by the medical experts to ensure the accuracy of the analytical models is necessary to gain credibility and acceptance of the solution for any meaningful clinical usage. Our team of medical experts from AIIMS New Delhi would engage in following activities for the success of the project.
Cloud will be used as an elastic platform for providing service endpoints to interface between the embedded POC system, client systems that are web-based on mobile applications and the scalable distributed storage system. The developed machine learning tools and packages will be deployed along with the service end points, making them available as a cloud based service. The service endpoints will act as a bridge between the heterogeneous components in the system and provide a means for platform independent communication.
Along with the communication provided to the POC device, a web-based client and a mobile/tablet will be provided to the users. When the patient creates an account, he needs to enter the basic details like age, sex (M/F), etc. The body weight and height will be measured by load cell and height scale. These values can be updated periodically and they are used for calculating measures like BMI and are also fed into the machine learning models as features.
At IIT Kharagpur, a web-based telemedicine system (called iMediX) has been developed for facilitating remote-health care services through its portal. This system will be customized for the specific need of uploading patient data, health parameters and bio-medical signal into its server. The system may be hosted over cloud. The data will be available to data-analytics tools and modules for aiding the decision making process of experts to intervene or not.