GEMINI: AI-Supported Breast Screening Finds Additional Cancers in an NHS Evaluation
Evidence: Prospective service evaluation with workflow simulations
Original study published: . This MedXchange brief was published on 2026-10-26.
Study Design and Methods
The Nature Cancer study evaluated Mia, a commercial mammography AI system, in 10,889 women at NHS Grampian. Routine double reading continued. Examinations flagged by AI but not recalled through routine screening underwent additional human assessment. Researchers examined 17 prespecified workflows. These combined a live additional-reading pathway with simulated triage arrangements, allowing comparison of different balances between cancer detection, recall and human reading workload.
Key Findings
Additional assessment identified 11 cancers: seven invasive cancers and four cases of ductal carcinoma in situ.
The primary combined workflow estimated a 10.4% relative increase in cancer detection, corresponding to one additional cancer per 1,000 women.
The same workflow projected up to 31% fewer human reads. This depended on the proposed reading and arbitration arrangements.
Clinical Significance
The additional cancers provide evidence that disagreement between AI and the routine reading process can reveal examinations deserving further assessment. The clinical result arose from AI directing attention to cases that were then assessed by people. The workload result addresses a different operational question. A modeled reduction in reading does not demonstrate that an entire department achieved that saving while maintaining its existing staffing, reporting and arbitration processes. Combining the two results into a single claim about autonomous screening would misrepresent the evaluation.
Clinical Interpretation
GEMINI makes the reading configuration part of the evidence. Its findings are most informative when the assessed workflow accompanies the performance figures. That allows radiologists to distinguish the contribution of an additional review from the projected effects of changing routine reading assignments. For clinical interpretation, detection, recall and workload should remain visible together. More detected cancers can be valuable, while the consequences for assessment activity and reading capacity remain separate questions.
Limitations
This was a service evaluation involving one product and one UK region, with live and simulated components. The workload projections are conditional on the modeled arrangements. The study does not establish a mortality benefit or show that the same results will occur with another product or screening population.
Explore the Work Behind Clinical AI
MedXchange Research Fellowship
Publish Clinical AI Research With Us
Our 10-week fellowship mentors physicians and researchers to build foundation models, analyze multimodal health datasets, and co-author peer-reviewed clinical studies.
Related Clinical AI Stories
AI-Derived Senescence Signature Stratifies Breast Cancer Prognosis and Immunotherapy Response
A novel senescence-related signature (SRS), developed through an integrative multi-omics and deep learning pathology approach, has demonstrated significant utility in stratifying breast cancer patients into distinct prognostic groups and predicting their response to immunotherapy. This study leveraged whole-slide pathology images, bulk transcriptomic profiles, single-cell RNA sequencing, spatial transcriptomics, and comprehensive clinical data to characterize senescence-associated features within the breast cancer microenvironment. The constructed SRS, derived from machine learning survival models applied to quantitative pathology features, consistently identified high-SRS patients with significantly poorer survival outcomes across internal TCGA-BRCA cohorts. The research further elucidated the cellular localization of senescence activity, predominantly in epithelial and endothelial cells, with fibroblasts playing a key role in mediating intercellular communication, particularly related to collagen. Crucially, tumors classified as low-SRS exhibited enhanced immune-cell infiltration, elevated immune-checkpoint expression, and more favorable predicted immunotherapy responses based on Tumor Immune Dysfunction and Exclusion (TIDE) and immunophenoscore (IPS) profiles. This integrated approach not only provides a powerful prognostic and predictive tool but also identifies specific senescence-associated biomarkers, such as STARD10, as potential therapeutic targets.
LLM Performance in PMOS/PCOS Information: DeepSeek-V4 Outperforms GPT-5.5 and Gemini 3.5 in Clinician-Rated Quality and Empathy
A recent study evaluated the information quality, readability, transparency, and perceived empathy of responses from three widely accessible large language models (LLMs)—DeepSeek-V4, GPT-5.5, and Gemini 3.5—to clinician-formulated questions regarding polyendocrine metabolic ovarian syndrome (PMOS), formerly known as polycystic ovary syndrome (PCOS). The study found significant differences in performance among the LLMs, with DeepSeek-V4 consistently receiving higher clinician ratings across multiple quality and empathy metrics. Thirty-three distinct English questions, crafted by clinicians and using established PCOS terminology, were submitted to each LLM. Two independent clinicians then rigorously assessed the 99 archived responses using standardized tools such as the DISCERN instrument for information quality, EQIP for transparency, and JAMA benchmarks. Readability was quantitatively analyzed using six different indices, while perceived empathy was subjectively rated. The methodological approach involved matched comparisons using Friedman tests, paired signed-rank permutation comparisons, and Holm correction to account for multiple comparisons.
Deep Learning Identifies p38α MAPK-MK2 Axis as Driver of Ovarian Cancer Chemoresistance via Nuclear Actin Remodeling
A novel deep learning pipeline, nFANet, has been instrumental in uncovering a critical mechanism by which ovarian cancer cells develop chemoresistance to platinum-based therapies. The study utilized nFANet for high-throughput analysis of nuclear F-actin assembly, revealing that platinum-based chemotherapy robustly induces this assembly. This process, facilitated by nuclear F-actin, is known to enhance DNA damage repair, thereby contributing to treatment failure in various cancers. Through a combination of nFANet and a subsequent kinome screening, the p38α MAPK-MK2 signaling pathway was identified as a key upstream regulator of chemotherapy-induced nuclear F-actin assembly. Mechanistically, this pathway promotes formin-mediated nuclear F-actin polymerization. Crucially, pharmacological or genetic disruption of the p38α MAPK-MK2 pathway significantly enhanced the sensitivity of ovarian cancer cells to platinum-based chemotherapy. These findings suggest a potential therapeutic strategy to overcome chemoresistance in ovarian cancer by targeting this specific molecular axis.