What happened
Koios Medical of Chicago makes Koios DS Breast, deep-learning software that gives radiologists a second opinion on whether a breast nodule on ultrasound is benign or malignant. AQuAS, Catalonia's agency for health quality and assessment, reviewed it for Spain's network of assessment agencies in 2024. It found four retrospective studies, three in the United States and one in Spain. In the largest, 15 readers classified 900 lesions with and without the software. Their area under the ROC curve was 0.87 with it and 0.83 without, a statistically significant difference. In another study, the commercial version left readers' area under the curve at 0.82, while two experimental versions raised it. In the Spanish study, following the software would have cut biopsies of 50 lesions rated BI-RADS 3 to 24, with none of the seven cancers missed. AQuAS found that a reader's sensitivity and area under the curve were equal or higher with the software. It warns that a high risk of selection bias may have inflated these results. Koios Medical funded one study, and AQuAS found no evidence on costs.
Sources
-
Assessment report, by authors with no tie to the system.
Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS), for the Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del SNS. Llinàs-Mallol L, Guzmán Aroca F, Estrada Sabadell MD, Vivanco-Hidalgo RM. Herramienta de apoyo en el diagnóstico ecográfico para la clasificación de nódulos mamarios. Ficha de evaluación de tecnologías sanitarias nuevas y emergentes. Barcelona: AQuAS, 2024. NIPO 133-24-102-2. Puntos clave.
Respecto a la validez diagnóstica de la tecnología, se ha observado que la sensibilidad y la abcROC de un evaluador es superior o igual con el uso de Koios DS Breast™, sin embargo, el elevado riesgo de sesgo relacionado con la selección de los pacientes podría haber sobreestimado los resultados de utilidad diagnóstica.
Details
AI system
- Workflow
-
The radiologist selected the nodule on two ultrasound images, and the software returned a malignancy category within three seconds. Readers saw the software's category after making their own assessment.
Follow-up
- Conditions for replication
-
The lesions resemble those studied, since accuracy fell for circumscribed, oval or anechoic nodules. Results hold in real clinical settings, where readers also know the patient's history and symptoms.
Entry details
- Sector
-
Healthcare (breast imaging).
- Purpose
-
Koios Medical built Koios DS Breast to classify breast nodules on ultrasound images by their probability of malignancy and give radiologists a second opinion.
30 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). Koios DS Breast helped readers classify breast lumps on ultrasound (AIBD-0929). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-0929.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)