Artificial intelligence (AI) in screening mammography can increase cancer detection. It can also significantly reduce workload by standing in as the second reader in traditional double reading settings.
However, what if AI could safely identify exams that may not require routine human review? New research seeks to answer that question.
This new, prospective publication from the ongoing Artificial intelligence in Breast Cancer Screening Program in Córdoba (AITIC) clinical trial, led by Spain’s Maimónides Biomedical Research Institute, sought to evaluate whether ScreenPoint Medical’s Transpara Detection algorithm can selectively reduce the need for human reading for a substantial portion of exams assessed as low risk.
The AI-based strategy resulted in radiologists reading only 36% of total exams (for a 63.6% workload reduction in screening readings) and also drove a 15.2% increase in the cancer detection rate (from 6.3/1000 to 7.3/1000) with a 14.8% increase in the recall rate from 4.8% to 5.5% (absolute difference of 0.7%) while maintaining a consistent positive predictive value of recall.
When analyzed by modality, no significant differences in cancer detection or recall rate were observed within tomosynthesis when comparing the AI strategy to standard double reading (while still achieving a 65.5% reduction in workload). In contrast, both cancer detection and recall increased within FFDM.
The authors note and address the ethical considerations of using only AI to review cases but suggest that “there are more cancers being missed by radiologists when not using AI for decision support in screening” (11 cancers missed when using Transpara compared to 54 cancers missed without the use of AI: and among those missed by the AI strategy, 91% were only detected by one of the two readers in the standard strategy, indicating their potential subtle appearance). Additionally, the AI strategy detected 10.1% more invasive carcinomas; 35% more carcinomas in situ; and a higher proportion of grade I invasive carcinomas, T1, and N0 invasive carcinomas than the standard strategy, without observing significant differences in the molecular profile
As the use of AI expands within the breast imaging practice and as evidence matures, this study adds to a growing body of literature that attests to AI’s potential benefits, yet it offers something unique and innovative in its approach and its intent. As the authors conclude, “AI triage and AI-supported screening that excludes low-risk mammograms from radiologist reading could be a safe and effective screening strategy that would allow for a substantial reduction in reading workload.”
Additional research may prove that to be the case.
Disclaimer: Transpara is intended for use as a concurrent reading aid for physicians interpreting screening FFDM and DBT exams, to identify regions suspicious for breast cancer and assess their likelihood of malignancy, in accordance with its FDA clearance and CE certification. Patient management decisions should not be made solely on the basis of analysis by Transpara. The views expressed herein reflect ongoing research, published evidence, and product development considerations. They do not describe or promote use of Transpara beyond its currently cleared and certified intended use. Any discussion of autonomous or rule-out applications represents future or investigational concepts and is not FDA cleared or covered by the current CE certificate.