It is now possible to predict survival of ovarian cancer patients, say researchers. They have created a new AI machine learning software that can predict survival rates and responses to treatment. The machine learning software focuses on patients with ovarian cancer.
The researchers, from Imperial College London and the University of Melbourne, wrote about their work in Nature Communications (citation below). The authors were Haonan Lu, Mubarik Arshad, Andrew Thornton, Giacomo Avesani, Paula Cunnea, Ed Curry, Fahdi Kanavati, Jack Liang, Katherine Nixon, Sophie T. Williams, Mona Ali Hassan, David D. L. Bowtell, Hani Gabra, Christina Fotopoulou, Andrea Rockall, and Eric O. Aboagye.
The authors’ machine learning software provides a more accurate ovarian cancer prognosis than current methods. After diagnosis, it can also predict the most effective treatments for patients.
The AI trial took place at Hammersmith Hospital in London, which is part of Imperial College Healthcare NHS Trust.
Prompter treatment for ovarian cancer patients
Their technology could help doctors administer the best treatment more quickly, the authors wrote. It opens a new avenue for more personalized medicine.
The researchers hope that their software can be used to stratify patients into groups based on the subtle differences in their cancer’s texture on CT scans rather than how advanced their cancer is or what type they have.
Lead author, Prof. Eric Aboagye, a Professor of Cancer Pharmacology and Molecular Imaging, at Imperial College London, said:
“The long-term survival rates for patients with advanced ovarian cancer are poor despite the advancements made in cancer treatments. There is an urgent need to find new ways to treat the disease.”
“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.”
AI may transform healthcare delivery
Regarding AI, Prof. Andrea Rockall, Clinical Chair in Radiology at Imperial College London, said:
“Artificial intelligence has the potential to transform the way healthcare is delivered and improve patient outcomes. 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.”
About ovarian cancer
Ovarian cancer usually affects patients (women) after the menopause. It can also affect people with a family history of the disease. It is the sixth most common cancer in women.
In the UK, there are approximately 6,000 new ovarian cancer cases annually. However, the long-term survival rate is only 30% to 40%. Survival rates are low because diagnosis typically occurs too late.
Early detection of this type of cancer would significantly boost survival rates.
Clinicians diagnose ovarian cancer in a number of ways. They may, for example, order a blood test to look for CA125, a substance that indicates the presence of cancer. The patient then undergoes a CT scan. The scan helps doctors determine how far the cancer has spread. It also helps them decide on the best treatment, such as chemotherapy or surgery.
The CT scans don’t, however, give doctors detailed insight into ovarian cancer patients’ likely overall outcomes. Neither do they help predict the likely effect of a therapeutic intervention.
The new machine learning software
The research team used TEXLab – a mathematical software tool – to identify the aggressiveness of tumors in CT scans and tissue samples from 364 ovarian cancer patients from 2004 to 2015.
The software examined the tumors’:
- genetic makeup,
- size, and
These four biological characteristics significantly influence overall survival.
The patients were then given an RPV (Radiomic Prognostic Vector) score. The RPV score indicates how severe the cancer is – from mild to severe.
The researchers compared the results with current prognostic scores and blood tests used by doctors to estimate survival.
Their software turned out to be up to four times more accurate than standard methods. Specifically, for predicting deaths from ovarian cancer.
The researchers also found that 5% of patients with high RPV scores had survival rates of less than 24 months. High RPV scores were also linked to poor surgical outcomes and chemotherapy resistance. This suggests that clinicians could use RPV as a potential biomarker to predict patient response to treatments.
The authors plan to carry out a larger study to determine how accurately the machine learning software can predict outcomes of drug therapies and/or surgery for individual patients.
What are AI and machine learning?
AI stands for Artificial Intelligence. The term refers to software that makes robots, computers, and other machines think like humans. It also makes them behave like humans. We have ‘natural intelligence,’ while smart machines have ‘artificial intelligence.’
Machine learning is an AI application that gives machines the ability to learn as they go along. In other words, learn from experience, without human intervention.
“A mathematical-descriptor of tumor-mesoscopic-structure from computed-tomography images annotates prognostic- and molecular-phenotypes of epithelial ovarian cancer,” Haonan Lu, Mubarik Arshad, Andrew Thornton, Giacomo Avesani, Paula Cunnea, Ed Curry, Fahdi Kanavati, Jack Liang, Katherine Nixon, Sophie T. Williams, Mona Ali Hassan, David D. L. Bowtell, Hani Gabra, Christina Fotopoulou, Andrea Rockall, and Eric O. Aboagye. Nature Communications, Volume 10, Article number: 764 (2019). DOI: https://doi.org/10.1038/s41467-019-08718-9.