My focus today is AI to determine cancer prognosis.
I have a special interest in breast cancer.
Knowing a patient’s prognosis is critical to offering optimal care.
I have written about the influences of artificial intelligence for those of us who work in medicine:
Artificial Intelligence May Render Some Medical Specialties Obsolete
Now, a groundbreaking new artificial intelligence model appears to outperform our current tests for predicting the chances breast cancer will spread or metastasize.
Join me as I explore this potential game-changer in the management of cancer.
“I never think of the future — it comes soon enough.”
― Albert Einstein
Cancer Characteristics Affecting Prognosis
Historically, the extent of breast cancer (stage) has been the leading prognostic factor.
The stage describes the amount of cancer in the body, where it is, and how far it has spread.
Early-stage breast cancer is less likely to come back (recur or relapse), so it has a more favorable prognosis.
The size of the primary cancer mattered, too.
Another important variable is biology:
- Is the cancer fed by hormones (such as estrogen or progesterone)?
- Does the malignancy appear aggressive under a microscope (grade)?
- Does the cancer cell have too many receptors for human epidermal growth factor receptor 2 (HER-2), a protein that controls cell growth and division?
How We Oncologists Currently Determine Prognosis
We can input the variables described above into a calculator that spits out projections of outcomes.
For example, the United Kingdom Predict calculator gives us the odds an individual will die of breast cancer in 5, 10, and 15 years.
Breast cancer survival prediction tool
Breast cancers develop at varying rates.
Some will grow quickly, while others grow more slowly.
We need good prognostic models for oncologists like me to optimize management for a given patient.
More recently, we got genomic testing.
Genomic Testing for Breast Cancer
Genomic tests (assays) examine a cancer tissue sample to determine certain genes’ activity.
The activity level of the genes can help predict how likely the cancer is to grow and spread (metastasize).
Here are examples of breast cancer genomic tests:
We routinely use Oncotype DX testing for the majority of patients with hormone-driven (estrogen or progesterone receptor positive) and not HER-2 positive.
The impact of genomic testing has been extraordinary, with one study reporting information for 479 patients with breast cancer.
Doctors in Ireland gave chemotherapy to 45 percent of patients.
Chemotherapy use changed in inverse proportion to the availability of the genomic assay.
Of those patients in whom Oncotype DX was utilized, 57 percent were spared chemotherapy.
The Future of Cancer Prognosis: AI to Determine Cancer Prognosis
A new company called Ataraxis AI has created a computer program that can predict how quickly a person’s cancer might spread.
To teach the program, they teamed up with hospitals and looked at many pictures of tumors and patient information.
They also made the program more accurate by creating several versions and combining their predictions, which helped eliminate mistakes.
AI versus Genomic Testing
They tested the program on information from 3,500 patients and reported this:
Artificial intelligence (AI) was much better at predicting cancer spread than current tests.
For estimating the progression risk of breast cancer, compared with tests such as Oncotype DX, AI was up to 30 percent more accurate.
Into the Future – AI to Determine Cancer Prognosis
The researchers offer that they look forward to improving their model’s accuracy.
I look forward to the developers making their AI model to healthcare providers as early as 2025.
Thank you for reading “AI to Determine Cancer Prognosis.”
