May 25th | 2018
Posted by: Kevin Goodwin, CEO
When it Comes to Ultrasound Based Devices More Information is Always Better… Especially When Using AI Methods
Ultrasound devices over the last 50 years have constantly improved, creating increasingly clear images in support of clinician decisions about a patient’s health status.
Ultrasound imaging involves using an ultrasound probe that sends sound beams into the body, retrieving an “envelope” of sound wavelets to then create an image. Over the years ultrasound scientists have figured out ways to increasingly “insonify” the body (sending more sound waves) and then extract far more information from the returning sound waves. This has resulted in much clearer pictures, in part, from the added “information” which enables ultrasound signal processing to perform more effectively by using all of those additional sound waves.
Almost 30 years ago, bladder scanners were created; and today conventional bladder scanners utilize the same “point & click” approach as when they emerged on the market in 1989.
Point & click became the original design for bladder scanning based on the belief that nurses would not do more than simply push a button. The point & click method simply pings a few cycles of sound waves in to the bladder aimed at computing a “good enough result” assuming the probe is properly positioned over the bladder itself. The small data sample limits the information available for computation of estimated bladder volume.
Now that AI methods have arrived for use on ultrasound devices, the need for more information has become paramount. Conventional bladder scanners have been stuck with point & click. These same devices that started the category nearly 30 years ago were never designed to optimize “quantity of information” but rather convenience, while also using older probe technology. (Why else would they require yearly calibration?)
In order to fully benefit from artificial intelligence methods (deep learning, convolutional neural networks, others) a device must acquire increased amounts of information so as to feed the advanced AI algorithms the data it requires.
The Uscan device introduced 2 years ago in May 2016 actually asks the nurse to fan across the bladder after first pointing the probe toward the pubic bone. Naturally, fanning results in more information being gathered from the patient’s bladder – as many as 256 slices are gathered by fanning. This increased package of sound waves is then sent to the ultrasound device for computation, creating a distinct advantage over the point & click method of insonification.
For AI to work well, clinicians must demand the maximum amount of information. They also must demand that bladder location & shape be known, as it is an “egg shaped three-dimensional object”. The more information the better.
Computational methods are advancing rapidly, as all AI are “data hungry”. Ultrasound devices love data because when you combine rich data sets with AI methods you can absolutely do a better job of delivering superior results for healthcare professionals.
It’s not surprising that conventional bladder scanner companies are arguing that “point & click” is all you need — and (even worse) — that point and click is AI.
This is particularly concerning because AI is a data-driven value-added methodology and, common sense suggests that “surveying the bladder via fanning” with an ultrasound probe that samples rapidly will gather far more data for the AI tools to compute.
The Uscan device also tracks movement information -from left to right- always knowing the angle of the probe itself. This results in the perfect combination of (256) slices of ultrasound information combined with angle data, yielding an information-packed 3D representation of the bladder; as much information as you could possibly gather from a bladder with an ultrasound-based tool.
When it comes to using ultrasound on the body “more sound in – more sound back” is the mantra. More information results in better computation and decision making.
With AI there is no question that more information is better and is capable of improving the day to day performance of care providers.