PulseHeal

By - Appani Kaushik

S P Mihir Achyuta

Chebrolu Vidya

Problem Statement

Novelty

To design a patient monitor system for hospitals which allows them to keep track of the patients with the use of IoT

• Most patient monitoring systems that are already available depend on signal acquisition and does not concentrate on signal processing with latest ML and deep learning techniques
• Almost all the hospitals have patient monitoring systems which cost a bare minimum of 35,000 INR and the sensors available in these are very precise. We intend to give a feature to use the existing systems by providing a way to connect to the network.

Abstract

Major issues with the current health monitoring systems is that they do not alert the available health care staff about the patients’ critical condition if anything untoward has occurred. They rely on wiring directly to the reception which is very complex as the number of beds or rooms keep on increasing and make it very difficult to repair if some problem has taken place.

In order to overcome the issue, we propose a system that relies on local network to monitor and send the patient data to the desk where the person monitoring is available and if required anywhere from the hospital to get alerts of the current situation of the patient. Also, to process the data for easier understanding of the signals such ECG. In order to make use of the existing systems in the hospital, we provide a method to integrate the existing system to the network.

Methodology

We will be using nodeMCU in the client side to interface sensors and to connect existing health monitors using UART protocol. The nodeMCU is connected to the local nodejs server. We are using a local server so that the patient data does not leak out of the hospital premises. For the front-end we will be using React and a dedicated server to perform the signal processing needed for the acquired sensor data.

Architecture

NodeMCU

Sensors

Patient Monitor (Existing)

Server

(Nodejs)

Front-end server

(Reactjs)

Data processing server

Hardware

A nodemcu is connected to the nodejs server which does a http post method every 10 seconds. A pulse sensor and a dht11 sensor is connected to find the pulse rate, SPO2 and the temperature of the patient. It has a feature to read from existing monitor and send data to the server.

Hardware

Front-end

The front-end is designed using react.js. It has an admin login page to view the patient data. It displays various patients in one screen and can show which patient condition is critical. It can also show live vitals of the patient in graphs.

Front-end

Front-end

Neural Network

• Convolution neural network to categorize into 5 types of heart beat - normal, supraventricular ectopic, ventricular ectopic, fusion and unknown
• Flask server to host the algorithm and return the type of heart beat

Business Model

• The hardware is installed with one time payment
• Software will be provided on a monthly/ yearly subscription basis
• The signal processing service will be add-ons depending on client

Target clients - Hospitals, diagnostic centers.