What question did this study set out to answer?

This research aimed to evaluate the effectiveness of Lumina 3D in enhancing diagnostic accuracy and reducing interpretation time for vessel occlusion and stenosis detection.

February 2, 2026

Abstract TP265: Reduced Interpretation Time, Improved Detection of High-Grade Stenosis and Occlusions using Lumina3D: A Multi-Reader, Multi-Case Study

Key Points

This research aimed to evaluate the effectiveness of Lumina 3D in enhancing diagnostic accuracy and reducing interpretation time for vessel occlusion and stenosis detection.
Conducted a multi-reader, multi-case study with 5 radiologists reviewing 20 CTA cases.
Utilized a randomized crossover design with cases of large vessel occlusions and high-grade stenoses.
Each radiologist reviewed cases twice—once with Lumina 3D and once without, with a one-month washout period.
Accuracy was assessed using mixed models, comparing reader results to expert consensus diagnoses.
Found a 9% improvement in diagnostic accuracy with Lumina 3D (85.6%) compared to without (76.1%).
Accuracy improvements varied by specialty: general radiologists improved by 12%, neuroradiologists by 6%.
Stenosis detection was 7% more accurate than occlusion detection.
Interpretation time decreased by 34 seconds per case, particularly benefiting general radiologists.

Abstract

Background: Timely detection of vessel occlusion and high-grade stenosis is crucial in stroke care. While CT angiography (CTA) source images are standard, interpretation can be time-consuming. 3D reconstructions improve anatomical clarity, but manual processing may take over 30 minutes. Lumina 3D TM by RapidAI automates this process, producing 3D head and neck reconstructions and curved planar reformats (CPRs) in minutes. This study evaluated Lumina 3D TM ’s impact on diagnostic accuracy and interpretation time. Methods: Five radiologists (3 general, 2 neuro) reviewed 20 head and neck CTA cases using a randomized crossover design. Cases included large vessel occlusions (8), distal medium vessel occlusions (16), and/or high-grade stenoses (12), defined as >75% luminal narrowing. Reference diagnoses were established by consensus of 3 expert neuroradiologists with full clinical access. Each reader reviewed all 20 cases twice—once with and once without Lumina 3D TM —separated by a one-month washout. The review order was randomized. Readers identified vessels with occlusions or stenoses, simulating clinical conditions. Mixed models assessed accuracy (percent agreement with reference), accounting for reader and scan variability. Results: The 20 cases included 24 occlusions and 11 stenoses across various vessels. With Lumina 3D TM , accuracy was 85.6% (95% CI: 73–98.5%) vs. 76.1% (95% CI: 63–89%) without it—an overall 9% improvement (p = 0.0004). Accuracy gains were independent of pathology type but varied by specialty: general radiologists improved 12%, neuroradiologists 6%. Stenosis detection was 7% more accurate than occlusion detection. Interpretation time decreased by 34 seconds per case (from 4.5 to 3.9 minutes; p = 0.027). General radiologists showed the greatest time savings (>1 minute). Conclusion: Automated 3D reconstructions via Lumina 3D TM improved diagnostic accuracy and reduced interpretation time, particularly for general radiologists. These findings suggest Lumina 3D TM may enhance stroke workflows and support faster, more accurate treatment decisions.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper