Related Experiment Video
Updated: Mar 7, 2026
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
23.4K
Can Radiologists Predict the Presence of Ductal Carcinoma In Situ and Invasive Breast Cancer?
Shadi Aminololama-Shakeri1, Chris I Flowers2,3, Christine E McLaren4
11 Department of Radiology, University of California Davis, 4860 Y St, Ste 3100, Sacramento, CA 95817.
AJR. American Journal of Roentgenology
|February 16, 2017
Summary
Radiologists' likelihood estimates accurately predict invasive cancer presence. These estimates also predict ductal carcinoma in situ (DCIS), though with less accuracy.
Area of Science:
- Radiology
- Oncology
- Medical Imaging Analysis
Background:
- Assessing malignancy risk in breast lesions is crucial for patient management.
- BI-RADS (Breast Imaging Reporting and Data System) categories guide diagnostic decisions.
- Radiologists' subjective likelihood assessments are integral to mammography interpretation.
Purpose of the Study:
- To evaluate if radiologists' estimated percentage likelihood of malignancy predicts histologic outcomes for ductal carcinoma in situ (DCIS) and invasive cancer.
- To determine the accuracy of these likelihood estimates in predicting specific cancer types.
Main Methods:
- Retrospective review of 250 BI-RADS category 4 or 5 breast lesions by 10 academic radiologists.
- Radiologists provided BI-RADS category, estimated percentage likelihood of DCIS/invasive cancer, and confidence ratings.
- Receiver operating characteristic (ROC) curves were generated to assess predictive performance.
Main Results:
- Area under the curve (AUC) values for invasive cancer ranged from 0.830-0.907, and for DCIS from 0.731-0.837.
- A 20% likelihood threshold predicted invasive cancer with 84% accuracy (sensitivity 82%, specificity 84%).
- A 40% likelihood threshold predicted DCIS with 80% accuracy (sensitivity 81%, specificity 79%).
Conclusions:
- Radiologists' estimated percentage likelihoods are effective predictors of invasive breast cancer.
- These likelihood estimates can also predict DCIS, but with lower accuracy compared to invasive cancer.
- Likelihood estimations offer a valuable tool for predicting histologic outcomes in breast imaging.
