Imaging at the Edge
Adoption of game-changing disruptive technologies in heavily regulated fields, such as healthcare, can prove slow, difficult and challenging. Learn how the use of Edge helps to overcome these challenges, transforming the world of medical imaging and providing innovative technology to healthcare professionals.
The concept of medical imaging began in 1895 with the invention of the x-ray by a German professor of physics, Wilhelm Conrad Roentgen. Since then, the range of technologies for medical imaging has expanded and today includes ultrasound, CT (computed tomography), PET (positron emission tomography), MRI (magnetic resonance imaging), CR (computed radiography), DR (digital radiography), and mammography.
Advancements in computing and digital transformation enabled replacing films with digital images, permitting multiplanar image reconstruction and the use of compute power to interpret scans. Medical imaging has evolved to become a key tool in assisting medical doctors to identify and often prevent health issues. Its wide adoption has led to an unforeseen challenge for the medical community – a global shortage of radiologists to analyse the increased volumes of data generated by these scans.
Paradoxically, this overwhelming volume of data presented the resolution. It provided the necessary critical mass for AI/ML algorithms to become an efficient tool capable of timely, accurate and tireless data processing while minimising human/operator error. AI/ML offered the means for a revolutionary paradigm shift in medical diagnostics.
The breakthrough came in combining the Edge and AI/ML enabling these algorithms to run on scanners or in close proximity, accelerating complex data analysis on-premises and improving clinical workflows.
Traditionally, images taken by a diagnostic machine, are sent to a centralised server for processing and analysis while radiologists wait for the results. In some cases, such as a stroke, any delay is critical as it may lead to irreversible consequences for the patient’s health, therefore establishing an accurate diagnosis in real-time is invaluable.
Today, comprehensive results from CAD (computer-aided diagnostics) enable medical professionals to detect early signs of disease increasing the quality of patient treatment and recovery and in some cases saving lives – something that had been unimaginable a few years ago.
Building and maintaining complex medical systems for highly regulated industries such as medical and healthcare is neither simple nor straightforward. Systems must be fully complaint with all relevant standards and regulations, such as FDA 510(k) and CE, patient data protection as defined by HIPAA and GDPR, and support security and hardening standards like FIPS 140-2, STIG, Common Criteria. System software changes, be it security patches, updates or upgrades, requires rounds of rigorous verifications and approvals to meet the requirements for safety and performance.
Deploying AI/ML on this node to perform initial, first level processing, has several benefits to hospital workflow and the quality of patient care. First of all, it helps radiologists to prioritise tasks, alerting them to critical cases requiring immediate attention. Second, the ability to get results in near real-time enables specialists, like surgeons, to efficiently perform complex procedures with guidance from scanners right in the operating theatre. Third, the ability to see scans in near-real-time helps scan technicians to ensure they are getting quality images.
In the middle of the 20th century, the brilliant British mathematician and computer pioneer, Alan Mathison Turing, introduced the idea of AI/ML. He envisioned a process whereby “computers would learn from experience as well as solve new problems through the use of guided principles”. Nearly 70 years later we are only scratching the surface of this possibility.