European journal of radiology · 2026
Financial impact of deep learning reconstruction in magnetic resonance imaging: experiences after widespread deployment.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Brix Mikael A K, Bode Michaela K, Järvinen Jyri, Keränen Riikka, Nikki Marko, Jussila Miro-Pekka, Räsänen Lasse, Korhonen Marko, Lammentausta Eveliina
Аннотация
PURPOSE: To evaluate the impact of deep learning reconstruction (DLR) on MRI productivity at a tertiary care academic hospital, and to validate a previously published Monte Carlo-based forecast of the productivity enhancement potential of DLR.
METHODS: Scanner log data were analyzed for two periods: pre-DLR adoption (January-October 2023, six scanners) and post-DLR (January-October 2025, five scanners). Examination and sequence durations were obtained from the Siemens Healthcare Teamplay platform. Observed changes were compared with capacity increases predicted by earlier Monte Carlo simulations. The impact of shortened scan durations across different scanner utilization levels was further evaluated using a simulation. Finally, four radiologists assessed image quality to identify potential clinical limitations of DLR.
RESULTS: Optimized scanners demonstrated a total reduction between 4.8 and 11.1 min (11.5-27.2%) in sequence duration and 5.0-10.7 min (9.5-21.2%) in total examination time. Despite operating with one fewer scanner in 2025 (six scanners before DLR adoption, 5 scanners after), the mean hourly throughput of the entire fleet decreased by only 6.4%, indicating improved productivity per scanner. Notably, this throughput was achieved while DLR deployment and protocol optimization were still in progress, underscoring the substantial productivity benefit even at a partial implementation stage.
CONCLUSIONS: DLR deployment was associated with improved MRI suite productivity, enabling nearly pre-DLR throughput despite operating with one fewer scanner. The results demonstrate the utility of DLR in improving MRI productivity and support the predictive accuracy of simulation-based health technology assessment. Nevertheless, unpredictable performance, particularly in neuroimaging, where artifacts in T2-weighted sequences and reduced quality in contrast-enhanced studies were observed, limits the applicability of DLR and underscores the need for rigorous quality assurance.
METHODS: Scanner log data were analyzed for two periods: pre-DLR adoption (January-October 2023, six scanners) and post-DLR (January-October 2025, five scanners). Examination and sequence durations were obtained from the Siemens Healthcare Teamplay platform. Observed changes were compared with capacity increases predicted by earlier Monte Carlo simulations. The impact of shortened scan durations across different scanner utilization levels was further evaluated using a simulation. Finally, four radiologists assessed image quality to identify potential clinical limitations of DLR.
RESULTS: Optimized scanners demonstrated a total reduction between 4.8 and 11.1 min (11.5-27.2%) in sequence duration and 5.0-10.7 min (9.5-21.2%) in total examination time. Despite operating with one fewer scanner in 2025 (six scanners before DLR adoption, 5 scanners after), the mean hourly throughput of the entire fleet decreased by only 6.4%, indicating improved productivity per scanner. Notably, this throughput was achieved while DLR deployment and protocol optimization were still in progress, underscoring the substantial productivity benefit even at a partial implementation stage.
CONCLUSIONS: DLR deployment was associated with improved MRI suite productivity, enabling nearly pre-DLR throughput despite operating with one fewer scanner. The results demonstrate the utility of DLR in improving MRI productivity and support the predictive accuracy of simulation-based health technology assessment. Nevertheless, unpredictable performance, particularly in neuroimaging, where artifacts in T2-weighted sequences and reduced quality in contrast-enhanced studies were observed, limits the applicability of DLR and underscores the need for rigorous quality assurance.