Artificial intelligence better predicts future risk of breast cancer

Some experts have warned of the drawbacks inherent in the spectacular progress achieved in the artificial intelligence (AI) because it has the ability to carry out tasks in a very short time and could replace the human being in many jobs. However, its advantages are indisputable, and proof of this is that a study has shown that artificial intelligence algorithms predict the risk of suffering breast cancer of a woman in the next five years.

The risk of a woman developing breast cancer is often estimated using tools such as the Breast Cancer Surveillance Consortium risk calculator (BCSC, for the acronym in English of the Breast Cancer Surveillance Consortium) of the United States that is based on the medical history and the information provided by the patient, such as her age, her family history of the disease, if she has had children, or if your breasts are dense, to calculate a risk score.

Good results have been obtained with this calculator, as demonstrated by a study published in JAMA Internal Medicine in 2019, but new research published in Radiology has revealed that AI has improved the prediction of breast cancer risk to five years. “Clinical risk models depend on the collection of information from different sources, which is not always available or collected,” he explained. Vignesh A. Arasu, research scientist and practicing radiologist at Kaiser Permanente Northern California, USA, and lead author. Recent advances in AI deep learning provide us with the ability to extract hundreds to thousands of additional mammographic features”.

AI could help deliver personalized, precision medicine

The researchers used data from mammograms 2D scans performed in 2016 at Kaiser Permanente Northern California centers that showed no visible signs of cancer. Of the 324,009 women who were screened to see if they met the eligibility criteria, 13,628 of them were selected for analysis. In addition, they also studied the 4,584 patients who were diagnosed with breast cancer in the five years after the 2016 mammogram and followed up all of them until 2021.

“AI is identifying both undetected cancers and breast tissue characteristics that help predict future cancer development”

The researchers divided the five-year study period into three time periods: interval cancer risk, or incident cancers diagnosed between 0 and one year; risk of future cancer, or incident cancers diagnosed between one and five years; and all cancer risk, or incident cancers diagnosed between 0 and 5 years.

To generate the breast cancer risk scores over a five-year period, we used two AI algorithms that were available to the researchers themselves and three that are already commercially available. The authors have highlighted that this is one of the few studies in which the algorithm is used to predict the risk of breast cancer years after in the mammogram the patient would have obtained a negative result. The results of the five algorithms were compared with each other and in comparison with other models.

“All of the algorithms we used outperformed the BCSC calculator in predicting 0-5 year risk of breast cancer,” said Vignesh A. Arasu, MD, adding: “This strong predictive performance over the five-year period suggests that AI is identifying both undetected cancers and breast tissue features that help predict future cancer development. There is something about mammograms that allows us to track breast cancer risk.”

The efficacy of AI stands out especially in the prediction of patients with a high risk of suffering from interval cancer, which is the one that is diagnosed between a screening examination evaluated as negative and the next screening appointment. In fact, when women who had a 10% higher risk of having this pathology were evaluated, the AI predicted up to 28% of cancers, while the BCSC calculator was only able to detect 21%. When both models were used, the results further improved the prediction of cancer.

Based on the results of the study, this technology offers “an accurate, effective and scalable means” of knowing the risk of breast cancer, since, as Arasu points out, in the traditional model only one source is used: the mammography. The researcher insists that this system should be generalized because it is also a tool that would help offer personalized and precision medicine: “AI for cancer risk prediction gives us the opportunity to individualize care for each woman,” concludes.


Leave a Reply