Machine Learning Driven Healthcare Monitoring System

For a UK-based medical service provider to improve clinical efficiency and patient care

Creation of a system that offers real-time human activity recognition within the clinical systems to provide better diagnosis & treatment for critical care patients

Capability

Analytics

Industry

Healthcare

Country

Key Features

The ML-based healthcare monitoring system offers actionable insights to help collate the recorded data (historical & current health data sets) from various connected devices along with the patient’s vitals.

Data collection

Readings are collected from various sources, including historical records, prescription, wearable patient sensors, and even equipment such as arterial blood gas (ABS) analyzer, echocardiography, and more

Improved efficiency

The simplified process helps convert varied data sets across different device protocols in a unified format for further analysis, reporting, diagnosis and, sometimes, treatment

Continuous monitoring

The system classifies the activities performed & recorded by different patients in a continuous sequence – while they are sleeping, sitting, being fed, walking, and more

Centralized health management

The AI-powered patient monitoring system offers real-time insights and alerts about health & performance, ensuring instant data interaction with actionable outputs

Challenges

No system in place to understand diverse device criticality and functionality

Manual & time-consuming to do analysis of historical & current data of the patients

No way to understand, diagnose and project next steps in patient-care in faster way

Need to offer real-time assistance for critical care patients

Solutions

We created ML-based data models for the connected devices to enable activity recognition through a secure & reliable patient health monitoring system. The model enables the medical care provider to record & analyze large data sets of patient health in real-time to make informed decisions.

Data set & deployment

The data set was divided into observation cycle windows of one second with a 50% overlap and further deployed into the AI system at scale on users’ devices

Smart data model

The model extracts the data for the mapping & analysis of the patient’s health status through connected devices, including a heart rate sensor, VO2, ECG, and more

Unified dashboard

The customized interface helps incorporate metrics beyond traditional activity identification models, and creates dashboard based on various parameters such as gender, age, illness, treatment group, and more