University of Aberdeen-Led Breakthrough in AI-Driven Breast Cancer Detection

The University of Aberdeen has spearheaded a groundbreaking study demonstrating how artificial intelligence (AI) can enhance breast cancer detection rates by 10.4 percent in real-world screening scenarios. Published on March 10, 2026, in the prestigious journal Nature Cancer, this research from the GEMINI project—conducted in partnership with NHS Grampian and the University of Glasgow—evaluates the Mia AI tool developed by Kheiron Medical Technologies (now part of DeepHealth Inc.). By integrating AI into routine mammography workflows, the study not only uncovers more invasive and high-grade cancers but also streamlines processes, offering hope for Europe's overburdened screening programs.

Breast cancer remains the most common cancer among women in Europe, with over 557,000 new cases reported in 2022 alone, and projections indicating a rise to 3.5 million globally by 2050. Early detection through mammography screening is crucial, yet current double-reading protocols by radiologists miss approximately 20 percent of cancers, leading to recalls for further tests where only one in five women receives a diagnosis. This University of Aberdeen initiative addresses these gaps head-on, positioning European universities at the forefront of health data science innovation.

How Mia AI Transforms Mammography Screening

Mia AI functions as a deep learning-based system trained on vast datasets of mammograms to predict malignancy risk. In the GEMINI study, it analyzed scans from 10,889 women aged 50-71 undergoing routine triennial screening in NHS Grampian between February and October 2023. While all participants received standard double human reading, Mia ran concurrently, flagging suspicious cases for arbitration when humans disagreed or missed potential issues.

The tool excelled at spotting subtle abnormalities—tiny tumors often invisible to the human eye. In live deployment, Mia identified 11 additional cancers from 1,345 arbitrated cases, including seven invasive tumors (six grade 2, one grade 3) and four ductal carcinoma in situ (DCIS) cases (three high-grade). Simulations of 17 workflow configurations further optimized outcomes, balancing sensitivity, specificity, and efficiency.

Key Findings: 10.4% Detection Boost Without Extra Recalls

The primary workflow—AI as an additional reader substituting or supplementing humans—yielded a 10.4 percent increase in cancer detection rate (one additional cancer per 1,000 screened), while reducing the overall recall rate by 0.8 percent. This meant fewer unnecessary biopsies and less patient anxiety, as positive predictive value improved.

• Cancer detection rate rose without inflating false positives.
• Workload for radiologists dropped by up to 31 percent (or 44 percent in triage modes).
• Time from screening to diagnosis notification plummeted from 14 days to just 3 days.
• Superior performance on invasive and high-grade cancers, enabling earlier, less aggressive treatments.

These results outperform prior European trials, such as Sweden's MASAI study from Lund University, which reported similar gains but higher workload reductions in select setups.

A Real Patient Story: Early Detection Saves Lives

Yvonne Cook, a woman in her 60s from Aberdeen, exemplifies the impact. During her 2023 routine mammogram, Mia flagged a minuscule Grade 2 tumor missed by radiologists. Recalled for further imaging, she underwent successful surgery and medication, avoiding chemotherapy. "I felt incredibly lucky," she shared. Without AI, detection might have waited three years or until symptoms appeared, risking spread.

Such cases underscore AI's role in shifting disease stages earlier, aligning with Europe's push for equitable screening where participation exceeds 80 percent in Nordic countries like Denmark and Sweden.

Photo by National Cancer Institute on Unsplash

Addressing Europe's Breast Cancer Screening Challenges

Across Europe, mammography programs invite women aged 50-74 every two to three years, yet radiologist shortages and rising caseloads strain systems. In the UK, over 2 million screens occur annually, but 20 percent of cancers evade detection. The University of Aberdeen study tackles this by demonstrating AI's adaptability—triage workflows prioritize high-risk cases, freeing experts for complex reads.

Comparable efforts include the University of Copenhagen's Danish trial (20 percent more detections) and Imperial College London's 13 percent boost. These university-led initiatives highlight a continental trend toward AI augmentation, potentially reducing mortality declines observed post-screening introduction.University of Aberdeen press release

Expert Perspectives from Aberdeen Researchers

Dr. Clarisse de Vries, lead author and former Aberdeen Research Fellow now at Glasgow, emphasized: "Our findings add high-quality evidence supporting AI, tailored to local needs amid workforce shortages." Professor Lesley Anderson, Chair in Health Data Science at Aberdeen, praised the novel trial design simulating real-world integrations.

Professor Gerald Lip, NHS Grampian Clinical Director, noted: "AI augments practice, delivering workload savings and burnout reduction in double-reading programs common across UK and Europe." These insights from Aberdeen's interdisciplinary team pave the way for policy shifts.

Implications for Radiologists and Healthcare Workforces

With Europe's ageing population driving breast cancer incidence up 38 percent by 2050, AI offers relief. GEMINI's 31-44 percent workload cuts could mitigate radiologist burnout, as flagged by the Royal College of Radiologists. Aberdeen's Grampian Data Safe Haven (DaSH) enabled secure analysis, a model for other universities.

For aspiring professionals, opportunities abound in health data science. Explore higher education jobs in AI and medical imaging at leading European institutions.

Future Outlook: From GEMINI to Nationwide Trials

The study bolsters the upcoming EDITH trial, a UK-wide NIHR-funded effort recruiting 660,000 women to test AI replacing one reader. Aberdeen, Glasgow, and NHS Grampian lead Scotland's component. Funded by NHS AI Award and Scottish Chief Scientist Office, this positions University of Aberdeen as a hub for translational research.

Broader European adoption could standardize AI via EU Cancer Screening Scheme, enhancing equity in lower-participation regions.

Photo by National Cancer Institute on Unsplash

University of Aberdeen's Role in European Health Innovation

Home to the Aberdeen Centre for Health Data Science, the university excels in AI applications, from breast screening to epidemiology. Collaborations with NHS and industry like DeepHealth exemplify academia-industry synergy. For researchers eyeing Europe, Aberdeen offers university jobs in cutting-edge fields.

This study not only advances detection but inspires curricula in data science and oncology across European colleges.

Actionable Insights and Next Steps for Stakeholders

Healthcare leaders should pilot site-specific AI thresholds, as GEMINI showed variability by workflow. Policymakers can leverage evidence for UK National Screening Committee recommendations. Patients benefit from faster, accurate results.

• Prioritize AI training for radiologists.
• Monitor biases via diverse datasets.
• Integrate with emerging tech like multi-cancer tests.

Visit higher ed career advice for paths in AI health research. For jobs, check higher-ed-jobs, university-jobs, and rate-my-professor.