A new study has shown that artificial intelligence (AI) software can help predict the risk of breast cancer recurrence in women who have had breast-conserving surgery for ductal carcinoma in situ (DCIS). The research looked at mammography results from 1,740 patients treated for DCIS and compared the accuracy of the AI tool with traditional clinical models.
The AI program analyzed mammograms and assigned risk scores. Women whose AI scores were above 73.5 percent had a much higher chance of cancer returning in the same breast. After five years, 4.13 percent of these women experienced recurrence, compared to just 0.86 percent for those with lower scores. At ten years, recurrence rates rose to 7.26 percent for higher AI scores and 3.72 percent for lower scores.
The AI software’s performance was similar to well-known clinical risk models, such as the Van Nuys Prognostic Index and the Memorial Sloan Kettering Cancer Center nomogram. For predicting cancer recurrence, the AI achieved an accuracy (AUC) of 70 percent, which was close to these traditional measures. The research team said this suggests AI-based imaging could be as reliable as clinicopathologic tools.
Interestingly, previous studies have linked certain mammogram features, such as calcification patterns or breast density, to a higher risk of recurrence. However, this new research did not find a connection between these features and the likelihood of DCIS returning. The authors believe AI’s objective assessment may be a strength compared to manual, subjective evaluations.
The researchers noted some limitations, including the use of one AI software, a relatively small number of cases, and data focused on Korean patients. These factors may make it harder to apply the findings to a wider population. Still, the study highlights that AI might play a valuable role in identifying patients who are at greater risk after breast surgery, which could improve follow-up care and treatment decisions in the future.
