AI can help detect breast cancer risk
![breast cancer](/resource/image/4214238/landscape_ratio2x1/320/160/e7813fe48da55e4e11a5d74d2e574e52/fG/breast-cancer-pix.jpg)
Some guidelines recommend that women undergo screening as late as age 50.
What you need to know:
- A recent study suggests that AI models can provide more accurate predictions by analysing hundreds of features in mammograms that are indicative of potential cancerous growth.
- Traditional risk assessments used by doctors rely on factors such as age, race, ethnicity, family history of breast cancer and previous breast tissue analysis.
Artificial intelligence (AI) has shown remarkable accuracy in identifying patients at the highest risk of developing breast cancer, surpassing the capabilities of the standard risk assessment currently used in health facilities.
A recent study suggests that AI models can provide more accurate predictions by analysing hundreds of features in mammograms that are indicative of potential cancerous growth.
Traditional risk assessments used by doctors rely on factors such as age, race, ethnicity, family history of breast cancer and previous breast tissue analysis.
However, "only about 15 per cent to 20 per cent of women who get diagnosed with breast cancer have a known risk factor such as family history of disease or previously having a breast biopsy," Dr Vignesh Arusu, a radiologist who was part of the study and specialises in cancer imaging at the University of California, San Francisco, said.
Dr Arasu and his team sought to explore how AI technology could assist in understanding future risk levels. In their study published in the journal Radiology, the researchers analysed data from 18,000 patients who had undergone mammograms in 2016 and were followed up until 2021. Approximately 4,400 of these participants developed breast cancer within the subsequent five years.
The AI models relied on mammograms that did not initially reveal any signs of cancer yet were able to identify certain patterns and features within the breast tissue that were linked to an increased risk of developing cancer. The team compared the performance of these AI models to the commonly used Breast Cancer Surveillance Consortium (BCSC) clinical risk model.
The AI models outperformed the BCSC model, particularly in predicting which patients were most likely to develop breast cancer within a year of their mammogram. Patients identified as having the highest AI risk scores, ranking in the 90th percentile, accounted for 24 per cent to 28 per cent of cancer cases, whereas the highest BCSC scores captured only 21 per cent of cases.
Dr Arasu suggests that incorporating AI alongside the traditional risk assessment model could enhance the accuracy of future breast cancer risk predictions.
While the study demonstrated promising results, it focused primarily on a non-Hispanic, white population. Dr Arasu emphasizes the need for further research to determine the efficacy of AI models across different races and ethnicities.
Also, it is unclear how the AI models may work for cancers of different severity.