Clinical Implementation and Validation of Automated Slice-to-Volume Reconstruction for Fetal MRI Using a Vendor-Neutral MONAI-Based Imaging Informatics Pipeline.
Authors
Affiliations (7)
Affiliations (7)
- Department of Radiology and Medical Imaging, Cincinnati Children's Hospital Medical Center, 3333 Burnet Avenue, Cincinnati, OH, 45229, USA. [email protected].
- Department of Radiology, University of Cincinnati College of Medicine, Cincinnati, OH, USA. [email protected].
- Fetal Care Center, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, USA. [email protected].
- Cincinnati Children's Artificial Intelligence Imaging Research (CAIIR) Center, Cincinnati, OH, USA.
- Department of Radiology and Medical Imaging, Cincinnati Children's Hospital Medical Center, 3333 Burnet Avenue, Cincinnati, OH, 45229, USA.
- Department of Radiology, University of Cincinnati College of Medicine, Cincinnati, OH, USA.
- Fetal Care Center, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, USA.
Abstract
This study aims to evaluate the clinical performance and operational reliability of a fully automated, vendor-neutral imaging informatics pipeline for fetal MRI slice-to-volume reconstruction using open-source Medical Open Network for Artificial Intelligence (MONAI) Deploy Express, DICOM routing, and PACS return. In this retrospective implementation study, consecutive singleton fetal MRI examinations performed at a tertiary fetal care center between October and December 2025 were processed using a MONAI Deploy Express pipeline triggered by DICOM routing. The pipeline performed automated series selection and motion-corrected SVR reconstruction for fetal brain and body imaging using routine single-shot fast spin-echo inputs. Reconstructed volumes were returned to PACS for radiologist review. Three fellowship-trained fetal radiologists independently scored reconstruction quality using a 5-point Likert scale. Primary outcomes included examination-level reconstruction acceptability, inter-reader reliability, failure modes, and operational time to PACS availability after transition to the finalized production workflow. Among 67 triggered examinations, 2 represented technical workflow events and were excluded from clinical performance analysis, leaving 65 examinations for image-quality evaluation. Mean gestational age was 28.9 weeks (range, 19-36 weeks) and most fetuses had structural abnormalities. Brain SVR was rated pass-or-higher in 59/65 examinations (90.8%) and body SVR was rated pass-or-higher in 57/65 examinations (87.7%). Inter-reader reliability was excellent for brain (ICC, 0.93) and body (ICC, 0.92) reconstruction quality assessments. Reconstruction failures were primarily associated with suboptimal or insufficient input stacks, severe fetal motion, amniotic fluid extremes, complex anatomic distortions, and early gestational age. In 57 consecutively processed examinations after finalized production integration, median time to PACS availability was 10 min 28 s for brain SVR and 11 min 50 s for body SVR. Automated fetal MRI SVR was operationally feasible using a DICOM-based, vendor-neutral deployment architecture integrated with routine PACS workflow. In this single-center, single-vendor cohort, the pipeline returned radiologist-acceptable reconstructions within a clinically practical timeframe. Multi-vendor and comparative reader studies are needed to establish geometric accuracy, generalizability, and diagnostic benefit.