Aidoc Brings Lifesaving AI Advancements to Medical Imaging on AWS

Some of the most dangerous injuries encountered by emergency-department (ED) doctors are invisible to the naked eye. While patients with open wounds and other obvious traumas may receive immediate care, patients with less apparent injuries like brain bleeds, spinal damage, or blood clots are sometimes left waiting—even while those injuries cause extensive and sometimes fatal damage.

The only way to detect internal injuries like these is by using medical imaging procedures such as computed tomography (CT), X-rays, and MRI scans, the results of which are analyzed by radiologists, medical doctors with advanced training in interpreting these images. Because of advances in these technologies, there are more and more of these scans per patient. This is good news, because it increases the likelihood of catching potentially deadly abnormalities that might previously have gone undetected. But as radiologists shoulder the resulting heavier workloads, many of them must make do with tools that aren’t up to the task.

Aidoc was founded to solve this problem by providing radiologists with advanced solutions that increase their efficiency and expedite patient care within their existing work environments. More specifically, Aidoc recognized that artificial intelligence (AI) and image-recognition tools could be used to augment radiologists’ expertise. The company’s always-on, AI-based, decision-support software analyzes CT scans to flag acute abnormalities, prioritize life-threatening cases, and expedite patient care. And it does all this on Amazon Web Services (AWS).

The Aidoc solution uses Amazon Elastic Compute Cloud (Amazon EC2) P3 instances to train machine learning (ML) models and execute inference processes; Amazon Simple Storage Service (Amazon S3) to store anonymized medical imagery for analysis; and Amazon Relational Database Service (Amazon RDS) to store image metadata.

"Since releasing our first FDA-cleared solution in September 2018, Aidoc has received FDA clearance for two more algorithms and has 12 more being used under FDA investigational device exemptions,” says Ariella Shoham, vice president of marketing for Aidoc. "The agility and development velocity that our AWS infrastructure enables have been major factors in helping us achieve so much success so quickly.”

Founded in 2015 by three Israel Defense Forces veterans with experience building AI solutions for military intelligence applications, Aidoc was born in the cloud. "Cloud computing is crucial for our work," says Guy Reiner, one of the company’s founders and its vice president of research and development. "Without the flexibility and scalability of the cloud, our offerings simply wouldn't be viable."

After comparing cloud providers, Aidoc decided to build its solution on AWS. "We selected AWS for its market leadership, range and maturity of services, partner ecosystem, and strong, flexible tools for security, scaling, and container management," says Reiner. "Because we work with hospitals, where data privacy is of the utmost importance, we also based our decision on the clear commitment AWS has to making its services HIPAA-eligible, as well as its extensive features and services supporting GDPR enablement in cloud environments."

Another key decision factor for Aidoc in selecting AWS was the availability of Amazon EC2 P3 instances. These instances provide powerful, highly scalable, parallel-processing capabilities well suited for the machine learning applications that are central to the company's core offerings. “One of the best things about running on AWS is always having access to the latest processor technologies, such as the new NVIDIA Tesla V100 GPUs, to help us make our model training as effective as possible,” says Reiner. "Considering all these factors together, we saw that AWS offered the best platform for the ambitious growth we were projecting."

On AWS, Aidoc's growth has easily kept pace with those initial expectations. "With the extremely fast development velocity AWS enables, we are getting new algorithms into production faster, bringing products to market faster, and helping treat patients and increase value for hospitals faster," says Reiner. "By using AWS, we are able to release algorithms targeting new pathologies every three months. That speed is unheard of in our industry, and it absolutely differentiates us."

Development velocity wouldn't mean much if the company couldn't get in the door at hospitals. According to Shoham, running on AWS has helped Aidoc win more than 300 customers in two ways. "First, we often encounter hospitals that are already running critical solutions on AWS, so we further strengthen credibility and trust when we say Aidoc runs on AWS, too. Second, because of the way we're able to architect our solution on AWS, we can assure hospitals that the data we bring to the cloud is anonymized, highly secure, and able to comply with even the strictest data-privacy rules."

Most important of all, Aidoc's solution, which to date has analyzed more than 3.2 million cases at more than 300 medical facilities around the world, is benefiting both patient care and radiologist well-being. At one major U.S. medical center, the Aidoc solution reduced patient ED visits by 59 minutes and overall hospitalization time by 18 hours, on average. At another similar medical facility, the Aidoc solution helped radiologists reduce the average turnaround time for CT scans of intracranial hemorrhages from 53 minutes to 46 minutes, or 13 percent.

“On AWS, Aidoc’s solutions are shortening patient wait times from days to hours for scheduled cases and from hours to minutes in the emergency department,” says Shoham. “It is very gratifying to have doctors who use our solution say things like ‘Today your solution helped us save a life.’ You can’t ask for more than that!”

Aidoc’s next steps include releasing a package of algorithms to support the work of oncology radiologists. “Our oncology package helps radiologists flag, compare, and perform volumetric measurement of kidney and liver lesions as well as lung and lymph nodules, reducing the bandwidth required for this demanding, time-consuming work,” says Reiner. “With a solution based on AWS, we know we will eventually be able to add value to almost every activity radiologists engage in."

To learn more, visit thinkwithwp.com/health.