Exploring Convergence Science Worldwide
Convergence thinking & policy
Mammography Screening Gets an Artificial Intelligence Upgrade
Conducted across four screening sites in Sweden, the study enrolled women aged 40-80 years, targeting both general screening and those with higher risk profiles. Participants were randomly allocated to either AI-supported screening or the traditional double reading process performed by radiologists. To ensure unbiased assessment, the radiologists remained masked to the study group allocation, while participants and radiographers were unaware of the reading protocol. The AI system employed in the trial, known as Transpara version 1.7.0, assigned a malignancy risk score to each examination, aiding in the triage of cases for single or double reading. Remarkably, the AI system demonstrated its proficiency by significantly reducing screen-reading workload, freeing radiologists to focus on cases requiring their expertise.
The safety analysis, conducted after enrolling 80,000 participants, also yielded promising results. The AI-supported screening exhibited a comparable cancer detection rate to standard double reading, with an impressive 44.3% reduction in screen-reading workload. Crucially, the study met the safety threshold, confirming a cancer detection rate above 3 per 1000 participants screened.
By harnessing AI's analytical prowess, mammography screening not only maintains its safety standards but also stands to achieve unprecedented efficiency. This pioneering approach could potentially revolutionise cancer detection, making it more accessible and resource-efficient without compromising accuracy. While this trial's focus was primarily on safety and efficiency, the long-term impact could extend to improving early cancer detection, leading to more effective interventions and improved patient outcomes.
Kristina Lång, Viktoria Josefsson, Anna-Maria Larsson, Stefan Larsson, Charlotte Högberg, Hanna Sartor, Prof Solveig Hofvind, Ingvar Andersson, Aldana Rosso
Mammography Screening Gets an Artificial Intelligence Upgrade
Conducted across four screening sites in Sweden, the study enrolled women aged 40-80 years, targeting both general screening and those with higher risk profiles. Participants were randomly allocated to either AI-supported screening or the traditional double reading process performed by radiologists. To ensure unbiased assessment, the radiologists remained masked to the study group allocation, while participants and radiographers were unaware of the reading protocol. The AI system employed in the trial, known as Transpara version 1.7.0, assigned a malignancy risk score to each examination, aiding in the triage of cases for single or double reading. Remarkably, the AI system demonstrated its proficiency by significantly reducing screen-reading workload, freeing radiologists to focus on cases requiring their expertise.
The safety analysis, conducted after enrolling 80,000 participants, also yielded promising results. The AI-supported screening exhibited a comparable cancer detection rate to standard double reading, with an impressive 44.3% reduction in screen-reading workload. Crucially, the study met the safety threshold, confirming a cancer detection rate above 3 per 1000 participants screened.
By harnessing AI's analytical prowess, mammography screening not only maintains its safety standards but also stands to achieve unprecedented efficiency. This pioneering approach could potentially revolutionise cancer detection, making it more accessible and resource-efficient without compromising accuracy. While this trial's focus was primarily on safety and efficiency, the long-term impact could extend to improving early cancer detection, leading to more effective interventions and improved patient outcomes.
Kristina Lång, Viktoria Josefsson, Anna-Maria Larsson, Stefan Larsson, Charlotte Högberg, Hanna Sartor, Prof Solveig Hofvind, Ingvar Andersson, Aldana Rosso
Leave a commentOrder by
Newest on top Oldest on top