The 3MT is a format whereby researchers explain their work in a tight three-minute time limit. There are no slides or props allowed! I took part in an informal 3MT event at university and found it a great experience, which helped me to condense my MSc project into the core principles at its heart. This is (roughly) what I said!
Who remembers the dress? You know, *that* dress. Was it blue and black? Or white and gold?
For anyone lucky enough to have avoided this phenomenon, it’s an example of an image that’s seen completely differently depending on who’s looking. No matter how much I might be convinced that I’m right and you’re wrong, it shows that human perception varies.
The dress debate may be best left far away on the internet… But what happens when it’s medical images we are looking at?
What if I were your doctor and the controversial picture was your scan? you’d want to be confident that I could be certain about what I could see.
And yet, one of the typical ways medical physicists check that X-ray machines are working is by taking a test picture and then counting the number of increasingly blurry, low-contrast features we can make out.
As you might have guessed, we might not always agree with each other.
If you can see 12 features one week and I only count 10 the next, is our equipment failing? Or have I just put my contact lenses in the wrong eyes again like I did last week?
I’m trying to get rid of this uncertainty by using a new measure of image quality that relies on computer software. Because, computers are a lot more consistent than we are. They aren’t as short-sighted as I am, they don’t get tired and they won’t change their mind depending on how sunny or dark the room is.
My computer software analyses images of a pretend tumour, and gives each image a detectability score. It considers contrast and noise, and simulates the response of human eyes to account for our vision.
I am using the scores to find out how to make lung tumours as visible as possible on CT scans.
CT scans are 3D X-rays. The scanner takes loads of X-rays from many angles around the patient, then combines them all into a 3D picture.
There are lots of different ways that it can do this combining. I’m going to scan a model lung tumour lots of times, using lots of different settings. I’ll then get my computer software to see how clearly it can see the tumour in each of these scenarios.
I’ll combine all the highest-scoring settings to make a theoretically optimum protocol for processing lung CT scans.
This gets rid of the subjectivity of using different people’s opinions over which protocol makes the best images. The computer will decide!
I’ll then test the new protocol against what we are currently using. Hopefully it will be significantly better!
Ultimately, it will help optimise lung CT scans, so that we can use less radiation, and still produce great-quality images, to enable doctors to see tumours clearly. Overall, it will be great to enable more effective assessments of lung cancer – which, unlike that dress, I’m sure we can all agree on.