Objectives of the service
Colorectal cancer is a worldwide disease. There are 1.3m new cases each year and 0.7m related deaths each year. Colonoscopy is an important screening tool to identify pre-cancerous and cancerous polyps, however, finding and diagnosing polyps is challenging. Studies have shown that doctors miss up to 25% of polyps and adenoma detection rates vary from 7-50%.
Artificial Intelligence can support doctors to locate and characterize polyps during the colonoscopy. To fulfil the potential impact AI can have on healthcare, it must be available to everyone from small clinics to large expert hospitals. This project explores how to make AI technology available to everyone through cloud deployment supported by satellite communications.
Additional activities are undertaken via the contract extension CCN#1. Such additional activities include: i) Clinical evaluation for the USA and regulatory pathway / approvals; ii) Integration of software into and demonstration of compatibility with endoscopy manufacturers existing endoscopic systems.
Users and their needs
Gastroenterologists and nurses endoscopists are under increasing pressure to deliver a higher quality and more diverse range of care in a funding and time constrained environments. Their performance is evaluated using key performance indicators (KPIs) which affects career progression and accreditation from the JAG (Joint Advisory Board on GI Endoscopy). Adenoma detection rate (ADR) is an important KPI.
Performance is highly correlated to training, experience (number of procedures) and frequency (number of procedures performed per week). Broadly speaking, in order to achieve good KPI’s the users have to identify all abnormal tissue and correctly characterize polyps for removal.
This remains challenging and recent reports have identified a wide variety in clinical practice.
Service/ system concept
EARTHSCAN is a cloud based artificial intelligence service that supports doctors to detect and diagnose polyps. Cloud computing allow the service to be more widely available. It is more scalable, flexible, powerful, cheaper and easier to upgrade. This is attractive for real-time healthcare systems as it also reduces the need to deploy/maintain hardware in hospitals. However, using the cloud requires a reliable IT network connection from the hospital to the cloud. The quality of hospital IT networks vary considerably between hospitals, regions and countries. They are under increasing pressure and demand is growing with more digital health applications and data.
EARTHSCAN uses a hybrid fixed and satellite communications solution to stream colonoscopy images from the hospital to the cloud for image analysis. The system architecture is shown below, outlies how data is communicated from the hospital to the cloud and the results are send back.
Space Added Value
Traditional access to the cloud relies on fixed or wifi communications. As this services is deployed in hospitals, it would rely on the hospital IT networks for the backbone of the communication. These systems vary greatly in the quality and capacity. They are under increasing strain and have variable workloads. Satellite communications and utilizing space operations compression algorithms enables us to establish a reliable and dedicated communication network for real-time health systems. Low cost satellite hardware and data service packages make this approach scalable and able to be deployed anywhere on earth.
Initial Pilot project has been completed. Pre pilot tasks of CCN#1, Factory and on-Site Acceptance Tests have been completed. Installation of system and Satellite at UCLH and testing of system in over 150 patients.