Credit: Pixabay.

Australian researchers developed an artificial intelligence tasked with predicting patient lifespans. The machine is surprisingly accurate at predicting which patient would die within five years. Its official accuracy rate was 69% or on par with a trained human oncologist and the researchers at the University of Adelaide behind it say an upgraded version is even better.

To train the machine to spot the right patterns that might predict lifespan, the scientist fed radiological chest scans belonging to 48 patients. The machine is based on a machine learning technique called deep learning which uses neural networks that mimic how the human brain works.

Subscribe to our newsletter and receive our new book for FREE
Join 50,000+ subscribers vaccinated against pseudoscience
Download NOW
By subscribing you agree to our Privacy Policy. Give it a try, you can unsubscribe anytime.

Lyle Palmer, an epidemiologist and one of the study’s authors, says his team embarked on this specific route because radiological images are a great resource. The more data you can feed into the machine’s algorithms, the more reliable its outputs, just like experience is important for making the right decisions in humans. Well, every year hundreds of thousands of radiological images are taken at every large hospital. Nowadays, most are well standardized and digitized which makes accessing them in a machine-friendly format is very easy. What’s more, many of these images contain bits of information that really no human can interpret — and not only the machine can, it can do so for thousands of instances at virtually the same time.

“In our recently published work, deep learning allowed us to explore the “hidden” features and patterns in CT images of the thorax that even expert humans are less able to decipher. We want to one day use this technology to predict the onset of chronic diseases such as diabetes, heart disease, and cancer before any symptoms are evident. As a proof-of-principle for our idea, we started off by looking at the much simpler outcome: 5-year mortality,” Palmer told ResearchGate. 

To compare their machine’s performance to a real human doctor, the researchers looked at clinical data and surveys that predict mortality published elsewhere. This is how they learn that their system had a similar accuracy as trained human doctors for gauging five-year lifespans, typically between 65 and 75 percent accurate. Adding more predictive information like age and sex will likely improve the result. In fact, since they published the paper, Palmer reports far better results.

“Our current models, which have built upon the foundational work described in our recent Scientific Reports paper, are capable of predicting 5-year mortality in a subset of patients being imaged with a thoracic CT scan in a hospital setting with around 80 percent accuracy. This is far better than a human doctor could do, and we are now thinking about how best to use this information in a clinical setting,” Palmer said.

“This sort of technology does not replicate what radiologists currently do, rather it provides an additional value to medical imaging studies. We envision that this approach could guide treatment decisions, potentiate preventative care, inform cohort selection for clinical research, and act as a more responsive biomarker for chronic diseases and aging,” he added.