What is the clinical evidence from this study?

Study design: Observational. Population: Atrial fibrillation (n=44). Intervention: Convolutional neural network (CNN) for automated LA segmentation vs. Semi-manual human contouring. Primary outcome: Dice score (DSC) for LA contouring accuracy.

What does this research mean for the field?

A convolutional neural network can fully automate left atrial segmentation and flow quantification in real-time phase contrast MRI for atrial fibrillation patients, achieving accuracy comparable to human experts while remaining resilient to heartbeat variability. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

April 29, 2025Open Access

Deep learning based automated left atrial segmentation and flow quantification of real time phase contrast MRI in patients with atrial fibrillation

Key Result

A convolutional neural network for automated left atrial segmentation and flow quantification in real-time phase contrast MRI achieved a Dice score of 0.90, comparable to human inter-observer agreement.

Study Design

Type

Observational (n=44)

Multicenter

Structured PICO

Does a convolutional neural network accurately automate left atrial segmentation and flow quantification compared to semi-manual analysis in patients with atrial fibrillation?

Population

44 patients with a history of atrial fibrillation (AF) or prior AF ablation (≥3 months prior), excluding those with prior left atrial appendage closures, clipping, or ligation.

Intervention

Fully automated left atrial (LA) segmentation and flow quantification using a 3D U-Net convolutional neural network (CNN).

Comparator

Semi-automated human observer analysis (magnitude intensity-based region growing algorithm with manual correction).

Outcome

Left atrial contouring accuracy (Dice score and Hausdorff distance) and flow measures (peak/mean velocity, net flow, and blood stasis).surrogate

A deep learning convolutional neural network can accurately and rapidly automate left atrial segmentation and flow quantification from real-time phase contrast MRI in patients with atrial fibrillation.

Main Result

Absolute Event Rate: 0.9% vs 0.93%

Limitations

Overestimated peak velocities in AI analysis for some patients due to excess contouring including the aorta or left ventricular outflow tract
Only one 2D RTPC plane placed in the left atrium parallel to the mitral valve was examined
Validated and tested using data only from a single site and scanner vendor, limiting generalizability
Small patient-level database for training, which may not account for total variation within all atrial fibrillation patients
Insufficient sample size to definitively determine the impact of heart rate variability on model performance

Abstract

Real time 2D phase contrast (RTPC) MRI is useful for flow quantification in atrial fibrillation (AF) patients, but data analysis requires time-consuming anatomical contouring for many cardiac time frames. Our goal was to develop a convolutional neural network (CNN) for fully automated left atrial (LA) flow quantification. Forty-four AF patients underwent cardiac MRI including LA RTPC, collecting a median of 358 timeframes per scan. 15,307 semi-manual derived RTPC LA contours comprised ground truth for CNN training, validation, and testing. CNN vs. human performance was assessed using Dice scores (DSC), Hausdorff distance (HD), and flow measures (stasis, velocities, flow). LA contour DSC across all patients were similar to human inter-observer DSC (0.90 vs. 0.93) and a median 4.6 mm 3.5-5.9 mm HD. There was no impact of heart rate variability on contouring quality (low vs. high variability DSC: 0.92 ± 0.05 vs. 0.91 ± 0.03, p = 0.95). CNN based LA flow quantification showed good to excellent agreement with semi-manual analysis (r > 0.90) and small bias in Bland-Altman analysis for mean velocity (-0.10 cm/s), stasis (1%), and net flow (-2.4 ml/s). This study demonstrated the feasibility of CNN based LA flow analysis with good agreements in LA contours and flow measures and resilience to heartbeat variability in AF.

Ask AI

Helpful

Bookmark

View Full Paper