New AI Model Helps Spot Normal Breast Screening Exams and Reduces DBT Workloads for Radiologists

Researchers from the University of Haifa (Haifa, Israel) conducted a study to evaluate the use of AI to reduce workload by filtering out normal DBT screens. DBT has higher diagnostic accuracy than digital mammography, but interpretation time is substantially longer. Nevertheless, the use of DBT is expected to show progressive growth worldwide, resulting in increased burden for radiologists and higher cost for screening programs. The use of AI models could help save time in the assessment of breast screening examinations and improve reading efficiency.

In the new study, the researchers proposed an AI model to detect cancer-free screening examinations that could be dismissed without consulting a radiologist to reduce workloads. The study included a large DBT screening data set with a substantial number of biopsy-proven examinations (1472 malignant cases and 2232 benign cases) collected from 22 clinical sites. In addition, their AI model examined both the DBT images and the clinical information with each DBT examination. The purpose of the study was to develop an AI model that could filter out normal DBT studies to reduce screening workloads while improving diagnostic accuracy. The researchers also performed a reader study to assess the effect of the use of an AI model in a simulated clinical workflow.

In the retrospective study, the AI model demonstrated the potential to reduce radiologists’ worklist by 39.6%, with improved specificity and non-inferior sensitivity. In a simulated workflow, the recall rate was reduced by 25%. When the team analyzed the AI false-negative findings, it found that almost 70% were occult at mammography. The researchers presented evidence of generalizability of the AI model, both to unseen patients and to unseen sites. AI performance was stable across all age groups, ethnicities, and body mass indexes, suggesting that AI may be widely applicable to diverse patient populations.

In the reader study, the readers had access to all information typically available during screening (such as previous studies and clinical information). The AI standalone performance was non-inferior to that of the mean reader. When worklist reduction for the mean reader was simulated, the specificity increased and recall rate decreased, with maintenance of non-inferior sensitivity. These findings strengthen the potential contribution of AI. Their analysis also showed that although AI performance was better in some metrics and non-inferior in others, its method of analysis is different from that of the human readers. This diversity provides additional support for AI’s potential to augment human decision making.

The researchers have theorized that trusting AI to perform radiologist’s work requires substantial evidence. The team believed that AI should be introduced into clinical practice gradually. Before AI is allowed to automatically interpret complex cases, it will first be used for tasks that are considered repetitive work, which was the approach taken in the study. According to the researchers, with time and with enough accumulated evidence, AI will be trusted in the same way as the results of automated blood tests are trusted.

Related Links:
University of Haifa