September 1, 2022 – It’s hard to figure out what the future path will look like for cancer patients.Considering a lot of evidence, such as the patient’s health and family history, the grade and stage of the tumor, and the characteristics of the cancer cells. But ultimately, the outlook comes down to health professionals analyzing the facts.
This can lead to “massive variation,” said Faisal Mahmood, Ph.D., assistant professor in the Department of Computational Pathology at Brigham and Women’s Hospital. Patients with similar cancers can end up with very different prognoses, some more accurate (or less) than others, he said.
That’s why he and his team have developed an artificial intelligence (AI) program that can result in more objective — and possibly more accurate — assessments.The purpose of the study was to judge whether artificial intelligence is a viable idea, and the team’s results have been published in cancer cell.
Because prognosis is key in determining treatment, higher accuracy could mean more treatment success, Mahmood said.
“[This technology] It has the potential to yield more objective risk assessments and thus more objective treatment decisions,” he said.
Building artificial intelligence
The researchers developed the artificial intelligence using data from The Cancer Genome Atlas, a public catalog of different cancer profiles.
Their algorithm is based on Histology (a description of the tumor and how fast the cancer cells are likely to grow) and genomics (using DNA sequencing to assess tumor at the molecular level). Mahmood points out that histology has been the diagnostic standard for more than 100 years, while genomics is increasingly being used.
“Both methods are now commonly used for diagnosis in major cancer centers,” he said.
To test the algorithm, the researchers selected the 14 cancer types for which the most data were available. When histology and genomics were combined, the algorithm gave more accurate predictions than either source of information alone.
Not only that, the researchers found, but the AI used other markers — such as a patient’s immune response to treatment — without being told to do so. That could mean AI can discover new markers we don’t even know about, Mahmoud said.
Although more research is needed – including large-scale testing and Clinical Trials – Mahmood believes this technology will one day be used by real-life patients within the next 10 years.
“Going forward, we’ll see large-scale AI models capable of taking data from multiple modalities,” he said, such as radiology, pathology, genomics, medical records and family history.
The more information an AI can consider, the more accurate its assessments will be, Mahmood said.
“We can then continuously assess patient risk in a computational, objective way.”