Using diagnostic images, artificial intelligence is better able to predict ovarian cancer survival than other current prognostic methods, a study suggests.
The study, “A mathematical-descriptor of tumor-mesoscopic-structure from computed-tomography images annotates prognostic- and molecular-phenotypes of epithelial ovarian cancer,” was published in Nature Communications.
A diagnosis of ovarian cancer is often confirmed with a computerized tomography (CT) scan, which creates a detailed image of a tumor using x-rays. These images can help clinicians make decisions about how far the disease has spread and what treatments might be effective — but they’re not very useful for predicting outcomes.
In the study, researchers employed a software tool called TEXLab to analyze CT images from 364 women with ovarian cancer between 2004 and 2015. The software assessed four basic aspects of a tumor — structure, shape, size, and genetic makeup — in order to create a novel prognostic indicator called “Radiomic Prognostic Vector” (RPV).
The program was fed algorithm data from patients treated in hospital and data available through online datasets. It split patients up into three RPV categories, namely high, medium, and low risk.
The researchers compared the patients’ RPV scores to blood tests and other prognostic scores currently used, and found their method was up to four times better at predicting deaths from ovarian cancer than standard methods.
Patients with high RPV scores were less likely to respond to conventional treatments, such as chemotherapy and surgery, and had the shortest average survival times — five per cent of patients with high RPV scores had a survival rate of less than two years. This suggests that RPV scores might be able to identify patients least likely to benefit from these treatments so that alternatives can be offered.
“Our technology is able to give clinicians more detailed and accurate information on how patients are likely to respond to different treatments, which could enable them to make better and more targeted treatment decisions,” Eric Aboagye, one of the investigators, said in a press release.
The researchers are hopeful that through further studies with more patient data, their software may be able to predict responses to treatment for individual patients, rather than for broad risk groups. Additionally, they speculated that the software might be used in the prognoses of other cancer types.
“Artificial intelligence has the potential to transform the way healthcare is delivered and improve patient outcomes,” said Andrea Rockall, Honorary Consultant Radiologist, at Imperial College Healthcare NHS Trust. “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.”