Related Experiment Video
Updated: Jul 2, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.5K
Diagnostic capabilities of artificial intelligence as an additional reader in a breast cancer screening program
Mustafa Ege Seker1, Yilmaz Onat Koyluoglu1, Ayse Nilufer Ozaydin2
1Department of Radiology, Acibadem Mehmet Ali Aydinlar University, School of Medicine, Istanbul, Turkey.
European Radiology
|February 22, 2024
Summary
This study shows artificial intelligence (AI) significantly improves mammogram screening accuracy and efficiency. AI as a triage tool enhances cancer detection, reduces radiologist workload, and enables earlier diagnoses.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Mammography screening programs aim for early breast cancer detection.
- Interval cancers and missed cancers remain challenges in screening.
- Evaluating artificial intelligence (AI) for early detection and workload optimization is crucial.
Purpose of the Study:
- To assess AI's early-detection capabilities in mammography screening over time.
- To evaluate AI's performance in detecting interval cancers and cancers from prior visits.
- To determine AI's impact on radiologist workload in different reading scenarios.
Main Methods:
- Analysis of 5136 mammograms from 4282 women in a 10-year screening program.
- AI software assigned scores; histopathology confirmed ground truth.
- Youden's index determined optimal AI threshold; bootstrapping evaluated workflow scenarios.
Main Results:
- AI achieved 89.6% AUC, with 72.38% sensitivity and 92.86% specificity at the optimal threshold.
- AI identified 51.72% of interval cancers and 50% of missed cancers.
- A hybrid triage workflow demonstrated a 69.5% workload reduction and 30.5% accuracy increase.
Conclusions:
- AI demonstrates high sensitivity and specificity for screening mammograms.
- AI effectively identifies interval and missed cancers, enabling earlier diagnoses.
- AI as a triage mechanism significantly reduces workload and improves screening efficiency.
