Evaluating the Accuracy and Efficiency of Virtual Reality Segmentation for Black-Blood Fetal Cardiac MRI.

BACKGROUND

CHD affects 1% of U.S. births annually, many requiring surgical intervention. Fetal echocardiography is the primary diagnostic tool but can be limited by fetal positioning or maternal anatomy, making black-blood contrast fetal cardiac MRI a valuable complementary modality.

OBJECTIVE

To evaluate a novel volumetric segmentation pipeline using virtual reality (Elucis, Realize Medical) applied to black-blood contrast fetal cardiac MRI scans for generating 3D CHD models.

MATERIALS AND

METHODS

Retrospective feasibility study.

POPULATION: Fetal cardiac MRI images from pregnant women as screening study for fetal CHD.

FIELD STRENGTH/SEQUENCE: High resolution 2D, T2-weighted single shot acquisitions obtained during fetal cardiac MRI. Acquisitions covering the entire fetus on 1.5T or 3T Philips Ingenia scanners in multiple orientations (~ 3-5 planes). Multiplanar acquisitions reconstructed to 3D dataset using slice-to-volume registration.

ASSESSMENT: Segmentation time, difficulty, and challenges were recorded per case. An advanced imaging cardiologist evaluated each 3D model for anatomical accuracy, diagnostic quality, required modifications, and structural visualization using standardized rating scale, with black-blood scans rated identically for comparison.

STATISTICAL TESTS: Descriptive statistics and paired t-test.

RESULTS

Visualized systemic veins/aortic arch in 100% of segmentations. Mean segmentation time was 4.7 min. Achieved diagnostic or minor-modification-only quality in 86.7% of cases. Segmentations showed significantly higher visualization scores than black-blood scans for systemic veins (3.8 ± 0.41 vs. 2.8 ± 0.99) and aortic/ductal arches (3.5 ± 0.57 vs. 3.2 ± 0.61); pulmonary artery (p = 0.546) and vein (p = 0.586) scores didn't differ significantly.

CONCLUSION

VR-based volumetric segmentation of black-blood fetal cardiac MRI produced diagnostic/near-diagnostic quality models in 86.7% of cases, with superior visualization of systemic veins and aortic/ductal arches versus black-blood scans alone.

EVIDENCE LEVEL: Level 3.

TECHNICAL EFFICACY STAGE: Stage 1.