Qure in India has developed AI technology to automate and speed up the analysis of medical CT Scans and X-rays for clinicians around the world, with key approvals in Europe and the US. Prashant Warier, CEO and co-founder of Qure details the hardware used for its medical artificial intelligence system.
The company is working with Astra Zeneca on an initiative to scan 5m people worldwide for early detection of lung cancer in Latin America and Asia. It has just completed an expedition to test the use of its AI with mobile X-Ray machines in extreme environments and has also worked with FujiFILM to detect TB in remote areas of South Africa.
CT scans and X-Rays are a critical part of the diagnostic pathway for many patients going through hospitals for treatment. It goes without saying the speed with which abnormalities are identified is essential to speedy diagnosis and treatment. In lung cancer where many patients are sadly not identified until it is late stage, early detection can have a dramatic impact on survival rates.
The machine learning technology developed by Qure allows incidental and progressive detection of abnormalities. In the past an X-Ray would be focused on analysing a specific area of concern, the AI technology can scan for multiple abnormalities at the same time, speeding up diagnosis and easing the burden on medical practitioners.
- Neural network detects Covid-19 in real time by Xray
- AI optical computer analyses Covid-19 X-rays
- Xilinx teams for cloud-based open source Covid-19 X-ray data
This progressive analysis of abnormalities was previously reliant on the clinician manually taking measures which is time consuming. It was also a 2D measurement. Now can be automated and delivers a more accurate 3D measurement making it more possible to identify concerning changes in an abnormality.
“Cloud computing drastically minimises the cost of implementing new technology. This allows components of our AI solutions to be installed on a low-end basic machine including PCs and smartphones,” said Warier.
“These components are built keeping in mind the need for patient privacy necessitating the use of edge computing. Consider edge computing to be a localised extension of a faraway supercomputer. We use edge computing to run simpler version of AI on the low-end systems to preserve patient privacy with protected and encrypted data at every level, both at source and in transit by ensuring that any data is de-identified before it leaves a client’s premises for cloud processing.”
“AI and imaging necessitate a large amount of computer power and given that healthcare providers have constrained resources, we are very conscious about working with hardware and infrastructure partners who offer extremely reliable systems that can be implemented in a cost efficient way,” he said.
“We don’t run AI on smart phone. The smartphone is just a viewer to see the outputs of AI. It runs very smoothly with the basic operating systems present in a personal computer or a phone. The algorithms have been built in a way that the demand for compute power is addressed through cloud based instances and the local devices act as the interface for image transfer and viewing the results of the images. This allows a user to upload a limited number of scans from simple devices accessible to them without compromising on the quality of AI reads.”
“Consequently, we have chosen to work with Amazon Web Services (AWS) and we actively focus R&D efforts on continually iterating and evolving our software to keep compute requirements to a minimum, despite rapid increase in product capabilities.
“One of the key drivers for our choice of cloud service provider is the need for high levels of security adherence. Hosting our solutions on servers compliant with local governance regulations allows us to use economies of scale and keep the costs to end users minimal.
“We are deployed on AWS cloud solutions across our sites. On AWS, we are using EC2 for heavy processing and easy scalability with better performance. It provides us with 99.8% service level agreement, which improves our performance while keeping the downtime to a minimum. We have also enabled automated backups and failovers in real time.
“When it comes to data security and privacy, our data is stored in S3 for better security, scalability, performance, and data availability. We also have RDS for database reliability. We are using CloudTrail & CloudWatch which monitor and record servers and account activity throughout AWS infrastructure. Moreover, we have AWS WAF, which is a web application firewall at perimeter level to secure our web apps & APIs against malicious traffic, web exploits, botnets, etc,” he said.
There are three main options for the applications to receive and process images for products. This can be cloud-based through Amazon, in a customer’s private cloud via an API or with the solution deployed on the customer’s local IT systems.
The company has published a handbook on considerations for the adoption of AI in healthcare. This looks at AI for radiology settings and is based on a workshop it ran at Intelligent Health AI in London with clinicians and policy makers.
Other articles on eeNews Europe
- Intel prepares for trillion transistor era shake up
- Imagination teams for RISC-V development platform for edge AI chips
- Opinion: Intel’s ‘smart capital’ is a warning from the past
- UK blocks purchase of Pulsic on security grounds
- 1400 RISC-V cores for on-chip machine learning