Can deep learning algorithms find breast cancer better than junior radiologists?
Deep learning (DL) is a type of artificial intelligence that learns patterns from large amounts of data. In breast cancer detection, DL algorithms are being tested to see if they can find tumors on mammograms and other scans as well as human radiologists. The short answer is that DL algorithms can perform as well as or better than junior radiologists in some settings, but they are not yet proven to be better than experienced radiologists. The evidence comes from studies comparing DL to radiologists of varying experience levels.
What the research says
A 2025 systematic review and meta-analysis of 13 studies with over 38,000 patients found that stand-alone DL algorithms for digital breast tomosynthesis (a 3D mammogram) achieved a pooled sensitivity of 0.88 and specificity of 0.74, with an area under the curve (AUC) of 0.89 2. This performance was not statistically different from all radiologists combined (AUC 0.89 vs 0.88) or from senior radiologists (AUC 0.89 vs 0.90) 2. However, DL showed significantly superior sensitivity compared to junior radiologists, meaning it caught more cancers 2. An earlier 2017 challenge study (CAMELYON16) found that the top DL algorithm for detecting lymph node metastases in breast cancer slides achieved an AUC of 0.994, outperforming a panel of 11 pathologists working under time constraints 9. In that study, the pathologists without time constraints performed similarly to the algorithm 9. Another 2025 study developed a DL model (iShape) to predict whether breast cancer has completely responded to chemotherapy in the lymph nodes, using ultrasound images. This model achieved AUCs of 0.88 to 0.92 across three external validation sets, showing strong performance 10. Overall, the research indicates that DL can match or exceed junior radiologists in specific detection tasks, but senior radiologists still perform at a similar level.
What to ask your doctor
- Could a deep learning algorithm be used to double-check my mammogram or ultrasound results?
- How does the experience level of the radiologist reading my scans affect the accuracy of my diagnosis?
- Are there any AI tools currently used at your facility for breast cancer detection?
- What are the limitations of AI in breast imaging that I should be aware of?
- If an AI algorithm finds something, does a radiologist still review it?
This question is drawn from common patient questions about this topic and answered using cited medical research. We do not provide individualized advice.