UW investigators co-develop successful new AI tool to map breast cancer
A team including UW Radiology postdoctoral researcher Anum Kazerouni, PhD, Professor and Executive Vice Chair of Radiology Habib Rahbar, PhD, FSBI, and Professor Savannah Partridge, PhD, have had a paper published in the Radiological Society of North American (RSNA) journal Radiology. The paper, “Cancer detection in breast MRI screening via explainable AI anomaly detection,” is notable for its “unprecedented accuracy” (1) in mapping out breast cancer anomalies on MRI results.
The team worked with investigators from Microsoft’s AI for Good Lab as well as Fred Hutch Cancer Center to develop an AI tool that can, “spot breast cancer in MRI scans more accurately than current digital methods, while also pinpointing exactly where suspicious tissue is located — a breakthrough that could make the sensitive screening tool available to more women.” (1)
Per an article from GeekWire on the model, “The system takes a novel approach by learning what normal breast tissue looks like, then flagging anything unusual, which is the opposite of how cancer-detection AI has typically been built. When it identifies potential cancer, it creates a visual heatmap showing radiologists precisely where to look.”
The AI model, “outperformed commonly used models in both high- and low-cancer prevalence situations.” (2) "This model really excels in imbalanced datasets, which is what we have in the scenario of breast screening, where the event of a cancer being detected is fairly rare," said Dr. Partridge in AuntMinnie. "It occurs in about 2% to 3% of exams, and that’s what we had here on our dataset that we trained the model on." (2)
Learn more about the study in the RSNA News article AI tool accurately detects tumors on breast MRI and in the Diagnostic Imaging piece, Explainable AI model for breast MRI shows effectiveness in high and low prevalence breast cancer datasets.
2) AuntMinnie.com: https://www.auntminnie.com/clinical-news/womens-imaging/article/15750646/explainable-ai-model-finds-breast-tumors-on-mri
