Frequently asked questions

BreastScreening-AI is a human-centered AI project for breast imaging. It aims to support radiologists with multimodal image review, AI-assisted second reading, explainable interaction and workflow measurement.

No. The product principle is that AI should strengthen, not replace, clinical judgement. A qualified clinician remains responsible for interpretation, diagnosis and patient management.

The project focuses on mammography, ultrasound and MRI because breast-imaging pathways often require more than one modality. Mammography supports population screening, ultrasound supports targeted characterization, and MRI supports selected high-sensitivity or complex-case assessment.

The public pages describe research software and progressive validation. They should not be read as regulatory clearance, clinical authorization or permission for routine diagnostic use.

The project addresses diagnostic uncertainty, avoidable callbacks, multimodal fragmentation, reader workload and the need for traceable human-AI decisions in breast imaging.

The intended workflow distinguishes the clinician's initial review, the AI recommendation, and the clinician's final decision. This preserves override, accountability and auditability.

False positives can create additional imaging, biopsies, patient anxiety and operational cost. The site discusses this as a workflow and economic issue, while avoiding claims that BreastScreening-AI has already reduced these costs in routine care.

Useful pilot measures include recall rate, benign biopsy volume, reader time, decision changes, disagreement, usability, workload, trust, false negatives, interval cancers where available, and implementation cost.

The evidence base includes peer-reviewed human-AI studies, multimodal imaging research, interaction-design studies, patent families, student-led academic work, and exploratory clinical-integration activities. Each result must be interpreted within its own study design and sample.

No. Results from controlled studies, exploratory pilots, proposal targets and business models should not be statistically combined unless the study designs, populations and endpoints support that analysis.

The curated bibliography is available on the Publications page. It includes project-relevant work rather than every publication by every associated author.

The claims register tracks numerical, clinical, regulatory and partnership statements so public content can remain accurate, source-backed and appropriately qualified.

The public website is informational. Contact forms, analytics, newsletter tools or support tools may process ordinary web and contact data. Medical images or patient records should not be submitted through public website channels.

Clinical research activities should use approved protocols, anonymization or pseudonymization where applicable, access controls, and institution-specific governance. Exact arrangements depend on the site, study and ethics approval.

No. Website content is not medical advice, diagnosis, treatment recommendation or emergency support. Patients should contact qualified healthcare professionals for clinical questions.

See the Terms, Privacy Policy, Cookie Policy and Disclaimer.

BreastScreening-AI can be described as a research-origin spin-off emerging from Instituto Superior Tecnico, Universidade de Lisboa. Public wording should remain aligned with the current intellectual-property and licensing status.

Potential collaborators include hospitals, imaging centers, research groups, PACS or imaging-technology teams, clinical-validation partners, regulatory specialists and health-economic evaluators.

Use precise categories such as clinical site, research collaborator, proposal consortium, technology environment or commercial partner. Avoid implying endorsement, deployment or signed partnership unless written evidence supports it.

Email info@breastscreeningai.com for collaboration, validation, documentation or business enquiries.