“Doctors and nurses are under a lot of stress to provide the best patient care. AI can go a lot way in assisting doctors during live procedures like colonoscopy. The Adenoma Detection Rate goes up, everyone benefits.”
“Roundtrip latency to the cloud can be high. That is not conducive for doctors using AI while performing critical procedures. AI needs to run on Edge devices. And only Avesha Smart Application Slices can 'edgeify' AI workloads & orchestrate them to bring real-time results.
The Procedure
20 Million Colonoscopies are performed in the US every year. During colonoscopy, a scope is inserted into the patient’s colon and the doctor then removes any abnormal growths (polyps), thus treating and preventing colorectal cancer. It brings immeasurable benefits to a patient, but the procedure can be challenging for the healthcare providers. Looping of the scope, patient discomfort, applying sustained pressure to the patient — all of these factors while detecting and removing polyps, documenting and reporting can add to the overwhelming conditions in a procedure room. Doctors & nurses perform these procedures for a straight 7–8 hrs in 30-minute intervals in a typical day.
AI, Edge and Avesha Smart Application Mesh
Application Smart Application Mesh is a technology platform that makes it easier to scale and operate cloud applications. The platform is an overlay to your existing infrastructure has intelligence and automation infused into it.
· It makes your application infrastructure smarter using AI and RL(Reinforcement Learning)
· It provides the surface area to easily move workloads around for hybrid, multi cloud and edge
The platform provides a simple declarative way of specifying the “intent” of an application with respect to what it wants to do with its workloads and data. For example, in the video inferencing for colonoscopy use case, the “application intent” is for data to be transferred to the right edge node capable of performing object detection using an AI model and directing the results to the procedure room monitors with minimal lag.
Avesha Edge Enablement: Video Edge Inferencing Solution
The primary goal of any healthcare provider is to provide the best care for patients without any life-critical errors.
Doctors and nurses are finding the help they need in
— artificial intelligence (AI) models running on edge servers
— streamlined workflows with NLP (Natural Language Processing)
— automation in generating reports with RPA (Robotic Process Automation)
The Avesha Video Edge Inferencing synchronizes all of the above workloads to provide real-time no lag results for critical procedures.
Thanks to these technologies, doctors now have an “extra pair of eyes” to assist them during polyp detection; improving accuracies and saving time — and potentially lives. AI Models that detect objects of interest inside the human body (in this case polyps in the colon) are complex Tensorflow models trained with hundreds of thousands of images from videos. Oftentimes, there are multiple models running in parallel that are trained in different classifications of polyps.
The Avesha Video Edge Inferencing platform synchronizes all AI model data flowing to and from the edge.
i) it provides extremely accurate object detection results with ultra-low latency
ii) it increases accuracy of results because the inferencing is done on video input as opposed to just images as input
ii) it combines outputs from multiple AI models in real-time to provide one synchronized output, which again improves accuracy and performance
iii) it improves efficiencies, streamline workflows & automates reporting with support for application intents like voice commands, NLP & RPA.
Real Time AI Detection Results
It is absolutely critical that there is no lag in the AI model’s assistance to the doctor during a live procedure. The doctor has to look through the monitor into the colon and identify polyps that may be subtle to distinguish — this is where a real-time AI assist can greatly benefit the doctor as a “second set of eyes”. Again, real-time results are critical. There can be no lag, the doctor cannot wait for the data from his scope to go to a model in the cloud and back. Thus it’s imperative for the AI model to analyze the scope video feed in an edge server in close proximity to the procedure room.
The Future: AI Assists for all procedures
Imagine a world where AI models are assisting doctors in every procedure room in a hospital/surgical center. There are AI models running for colonoscopy procedures, for biliary lesion detections, for cardiac procedures, for pulmonary procedures and more There is the need for an overlay technology that can sit above a distributed edge network & segment each AI model application traffic ensuring no-lag detection, ultra-low latency, security & privacy. Each AI model and its related application intents will be orchestrated by its own Application Slice created on the Avesha Smart Application Mesh platform. Only the Avesha products and solutions can bring order and efficiency to such a simultaneous activity of real-time AI assisted procedures.
Medical AI and Edge Computing are the future and Avesha Video Edge Inferencing is just the innovative platform needed to operationalize an end-to-end solution.
Avesha along with Verizon and AWS have successfully demonstrated Video Edge Inferencing for colonoscopy AI on AWS Wavelength & Verizon MEC nodes. For details please see https://www.youtube.com/watch?v=5uikuXvXGOA&t=10s
