Dr Marie McIlroy
Honorary Secretary of the Irish Association for Cancer Research. Senior Lecturer, RCSI University of Medicine and Health Sciences
Generative artificial intelligence (AI) and large language models are transforming every facet of society — cancer research is no exception.
At the recent Irish Association for Cancer Research (IACR) 2025 meeting in Belfast, the session ‘AI in Cancer Research’ showcased how cutting-edge AI technologies are reshaping approaches across cancer biology.
Discussions ranged from mapping genetic vulnerabilities in cancer cells to understanding interactions between cancer drivers and the tumour immune microenvironment. In a standout moment, Dr Fredrik Strand, Consultant Radiologist at the Breast Imaging Unit at Karolinska University Hospital, presented about a clinical study leveraging machine learning to improve breast cancer outcomes.
ScreenTrustMRI trial
The study1 (ClinicalTrials.gov registration: NCT04832594) addressed a critical challenge: breast cancers are often missed by routine mammography, particularly in women with dense breast tissue. Breast density is the proportion of fibroglandular tissue compared to fatty tissue within the breast. High breast density poses a challenge for standard mammogram interpretation because dense tissue appears white on the image, similar to many tumours, making it difficult to visually detect small or suspicious lesions. While MRI offers superior accuracy for early cancer detection, its high cost limits its use in standard screening programs. To bridge this gap, the team developed AISmartDensity, an AI tool that reviews mammograms to identify women at high risk of undetected cancer, recommending supplemental MRI for these cases.
The results were striking. The detection rate — 64.4 breast cancers per 1,000 MRI exams — is almost four times higher than the DENSE2 trial, which relied on traditional breast density measures.
Although widespread MRI referral may not be feasible in all healthcare systems, AISmartDensity offers additional value by flagging suspicious mammograms for radiologist re-review, enabling earlier intervention for high-risk patients.
The detection rate — 64.4 breast cancers per 1,000 MRI exams —
is almost four times higher than the DENSE2 trial, which relied on traditional breast density measures.
AI’s promise for precision oncology
This study underscores the transformative potential of AI in cancer detection, offering just a glimpse of what machine learning can achieve in advancing precision oncology. A recent review in Nature Cancer3 highlights three areas where AI is poised to deliver the greatest impact: cancer prevention and diagnosis, optimisation of existing treatments and acceleration of new therapy development. Collectively, these innovations demonstrate how AI isn’t only enhancing current clinical practice but also redefining the future of cancer research and treatment.
