Science

AI system four times better at predicting ovarian cancer patient survival than other methods

AI system four times better at predicting ovarian cancer patient survival than other methods
A new AI system can effectively predict those at the highest risk of dying from ovarian cancer with greater accuracy than existing methods, such as CT scans
A new AI system can effectively predict those at the highest risk of dying from ovarian cancer with greater accuracy than existing methods, such as CT scans
View 1 Image
A new AI system can effectively predict those at the highest risk of dying from ovarian cancer with greater accuracy than existing methods, such as CT scans
1/1
A new AI system can effectively predict those at the highest risk of dying from ovarian cancer with greater accuracy than existing methods, such as CT scans

An international team of researchers, from Imperial College London and the University of Melbourne in Australia, has demonstrated a new AI system that can effectively predict survival rates from ovarian cancer better than any current conventional method available to doctors.

"The long-term survival rates for patients with advanced ovarian cancer are poor despite the advancements made in cancer treatments," explains Eric Aboagye, lead author on the new research. "There is an urgent need to find new ways to treat the disease."

The five-year survival rate for ovarian cancer sits at around 40 percent and this is primarily due to challenges in detecting the disease at an early-stage. Doctors currently have several tools in their arsenal to diagnose the cancer, including CT scans and blood tests. CT scans in particular offer doctors a picture of the cancer that subsequently guides treatment decisions.

The new AI tool has been developed to offer doctors a better guide to how best treat a specific patient. The machine learning algorithm was trained on 10 years' worth of CT scan and tissue sample data from 364 women. Four tumor characteristics were evaluated retrospectively by the system: structure, shape, size and genetic makeup. The system was then able to give each patient a disease severity rating called a Radiomic Prognostic Vector (RPV).

The higher the RPV score the worse the survival rate, with those subjects receiving the highest score demonstrating poor chemotherapy response and an increased likelihood of dying from the cancer within two years. Overall the system was found to be four times more accurate at predicting patient survival rates compared to conventional diagnostic methods.

"Artificial intelligence has the potential to transform the way healthcare is delivered and improve patient outcomes," says study co-author Andrea Rockall. "Our software is an example of this and we hope that it can be used as a tool to help clinicians with how to best manage and treat patients with ovarian cancer."

The efficacy of the system will need to be verified in larger patient populations before it is more broadly rolled out, however the researchers suggest it can confidently produce RPV scores within just a few minutes of examining a patient's dataset. One significant benefit of the prospective system is its ability to predict a patient's response from standard surgical or chemotherapy treatments. This prediction could help clinicians better select ovarian cancer patients most suited to experimental or trial treatments at an early stage when the treatment may be most successful.

The new research is published in the journal Nature Communications.

Source: Imperial College London

No comments
0 comments
There are no comments. Be the first!