Researchers wanted to see if MRI scans could help predict HER2 protein levels in breast cancer tumors before surgery. HER2 status helps guide treatment decisions. They studied 375 young breast cancer patients under age 40 from two medical centers, using their preoperative MRI scans to build computer prediction models.
The study compared different prediction approaches. One method analyzed the entire tumor, while another looked at different areas within the tumor. The models that combined information from different tumor areas performed best at distinguishing between HER2-negative and HER2-positive tumors. For a more detailed task of separating zero HER2 from low HER2, combined models also showed the highest accuracy.
This was a retrospective study, meaning researchers looked back at existing patient data rather than testing the approach in real time. The performance numbers reported came from the training phase of the models, not from independent validation. No safety concerns were reported because this was an analysis of existing scans, not a new treatment.
Readers should understand this is early research exploring a possible future tool. The study shows MRI-based analysis might one day help predict HER2 status non-invasively in young patients, but much more research is needed before doctors could use this approach in practice.