Although the night team has been functioning like a well-oiled machine, the summer heat combined with the constant stream of patients has resulted in two very long nights. Amongst the chaos of resus a reg was trying to insert a central line which was guided by ultrasound. Ever since I saw conservation of momentum used in mathematical models of ultrasound, I’ve mildly fallen in love with it.
“Did you know they’re using machine learning in ultrasound?”
I said with mild enthusiasm. The reg who is usually friendly, approachable, and nice to work with quickly replied:
“Yes but the computer won’t insert the line”
I was mildly taken aback as I was not implying that this would replace him. Yes machine learning can be used to make predictions and facial recognition software is a result of machine learning but machine learning is used for so much more. In short, a machine learning algorithm tweaks the parameters of an equation based on the data it is being fed. This is helpful for signal processing and has led to publications using machine learning in order to reduce noise in ultrasound images.
The reactions from clinicians generally seems to be absolute. Some completely dismiss it as if nothing they do can be automated whilst others get defensive, quickly pointing out the computer’s flaws. In reality, machine learning will automate some things, but it will also make more things possible. It’s ability to quantify and measure effect size of variables implies that it could increase the amount of decision making for clinicians in the future. Another area where technology has done this is Google. Internet searching has reduced the skill and effort needed to find general facts, but it has also empowered the average person to make many more decisions about their life. However, if the person is not very smart, Google isn’t going to stop them making stupid decisions. The same goes for data and probabilities presented by machine learning algorithms. The clinician understands the practicalities of implementing choices, not the computer.