Cancer is a huge killer in Scotland. The currently incurable disease affects not only the person diagnosed, but their loved ones, too.
Research is constantly underway to improve cancer survival rates, develop new treatments, and find ways to prevent cancer. And now, Johns Hopkins University in Maryland, USA is going one step further with the use of AI.
Radiologists have already began using AI-based computer vision models to help speed up the laborious process of analysing medical scans.
Previous abdominal organ datasets were compiled by radiologists manually identifying and labelling individual organs in CT scans, requiring thousands of hours of human labour.
This is where the team at John Hopkins have a “solution” – AbdomenAtlas, the largest abdominal CT dataset to date, featuring more than 45,000 3D CT scans of 142 annotated anatomical structures from 145 hospitals worldwide.
Using AI algorithms to dramatically accelerate this organ-labelling task, the Hopkins-led team completed in under two years a project that would have taken humans alone over two millennia.
“Annotating 45,000 CT scans with six million anatomical shapes would require an expert radiologist to have started working around 420 BCE – the era of Hippocrates – to complete the task by 2025,” says lead author Zongwei Zhou.
By repeating a process – AI prediction followed by human review – researchers claim to have significantly accelerated the annotation process, achieving a 10-fold speedup for tumours and 500-fold for organs.
This approach enables the team to expand the scope, scale, and precision of their dataset without overburdening radiologists, resulting in what the team says is the largest fully annotated abdominal organ dataset in existence.
What’s more, they are adding more scans, organs, and both real and artificial tumours to help train new and existing AI models to identify cancerous growths, diagnose diseases, and even create digital twins of real-life patients.
“By enabling AI models to learn more about.. [things like] tumour identification… we have made AI perform similar to the average radiologists in some tumour detection tasks,” said one of the authors, Wenxuan Li.
There is one caveat. Despite AbdomenAtlas’ success, the dataset only accounts for 0.05 per cent of the CT scans annually acquired in the US. They are calling upon other institutions to help fill in the gaps.
Abdominal cancer in Scotland includes stomach cancer, oesophageal cancer, and other upper gastrointestinal (GI) cancers. Early diagnosis can improve the chances of survival.
Don’t miss the latest news from around Scotland and beyond – Sign up to our newsletterhere.