Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Two-Stage Deep Learning Framework for Quality Assessment of Left Atrial Late Gadolinium Enhanced MRI Images

K M Arefeen Sultan, Benjamin Orkild, Alan Morris, Eugene Kholmovski, Erik Bieging, Eugene Kwan, Ravi Ranjan, Ed DiBella, Shireen Elhabian

TL;DR

The study tackles automated quality assessment of left atrial LGE-MRI images for fibrosis quantification, addressing limited labeled data. It introduces a two-stage framework: (i) left atrium detection to localize relevant slices via segmentation, and (ii) quality assessment using a CNN-based encoder with Baseline QA, Multi-Task QA, and supervised contrastive pretraining strategies. Contrastive pretraining and multi-task learning yield the most robust improvements in F1-Score and Specificity under data scarcity, with the QA-Contrastive approach performing best on held-out scans. The method enhances diagnostic reliability and efficiency in fibrosis assessment, supporting standardized, scalable LGE-MRI QA for improved treatment planning in atrial fibrillation.

Abstract

Accurate assessment of left atrial fibrosis in patients with atrial fibrillation relies on high-quality 3D late gadolinium enhancement (LGE) MRI images. However, obtaining such images is challenging due to patient motion, changing breathing patterns, or sub-optimal choice of pulse sequence parameters. Automated assessment of LGE-MRI image diagnostic quality is clinically significant as it would enhance diagnostic accuracy, improve efficiency, ensure standardization, and contributes to better patient outcomes by providing reliable and high-quality LGE-MRI scans for fibrosis quantification and treatment planning. To address this, we propose a two-stage deep-learning approach for automated LGE-MRI image diagnostic quality assessment. The method includes a left atrium detector to focus on relevant regions and a deep network to evaluate diagnostic quality. We explore two training strategies, multi-task learning, and pretraining using contrastive learning, to overcome limited annotated data in medical imaging. Contrastive Learning result shows about $4\%$, and $9\%$ improvement in F1-Score and Specificity compared to Multi-Task learning when there's limited data.

Two-Stage Deep Learning Framework for Quality Assessment of Left Atrial Late Gadolinium Enhanced MRI Images

TL;DR

The study tackles automated quality assessment of left atrial LGE-MRI images for fibrosis quantification, addressing limited labeled data. It introduces a two-stage framework: (i) left atrium detection to localize relevant slices via segmentation, and (ii) quality assessment using a CNN-based encoder with Baseline QA, Multi-Task QA, and supervised contrastive pretraining strategies. Contrastive pretraining and multi-task learning yield the most robust improvements in F1-Score and Specificity under data scarcity, with the QA-Contrastive approach performing best on held-out scans. The method enhances diagnostic reliability and efficiency in fibrosis assessment, supporting standardized, scalable LGE-MRI QA for improved treatment planning in atrial fibrillation.

Abstract

Accurate assessment of left atrial fibrosis in patients with atrial fibrillation relies on high-quality 3D late gadolinium enhancement (LGE) MRI images. However, obtaining such images is challenging due to patient motion, changing breathing patterns, or sub-optimal choice of pulse sequence parameters. Automated assessment of LGE-MRI image diagnostic quality is clinically significant as it would enhance diagnostic accuracy, improve efficiency, ensure standardization, and contributes to better patient outcomes by providing reliable and high-quality LGE-MRI scans for fibrosis quantification and treatment planning. To address this, we propose a two-stage deep-learning approach for automated LGE-MRI image diagnostic quality assessment. The method includes a left atrium detector to focus on relevant regions and a deep network to evaluate diagnostic quality. We explore two training strategies, multi-task learning, and pretraining using contrastive learning, to overcome limited annotated data in medical imaging. Contrastive Learning result shows about , and improvement in F1-Score and Specificity compared to Multi-Task learning when there's limited data.
Paper Structure (13 sections, 5 equations, 3 figures, 2 tables)

This paper contains 13 sections, 5 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methods
  4. Left Atrium Detection Stage
  5. Quality Assessment Stage
  6. Baseline QA
  7. Multi-Task QA
  8. Pre-training using supervised contrastive learning
  9. Results
  10. Dataset
  11. Data Preprocessing & Augmentation
  12. Summary of Experiments
  13. Conclusion

Figures (3)

  • Figure 1: Architecture of our model.
  • Figure 2: Result of HiResCAM hirescam between our Baseline QA and Multi-Task QA model output on $2$ different test scans. We show the critical cases where our QA model fails to focus on (blue color) the relevant areas for scoring, whereas, on the contrary, the Multi-Task QA model is able to do so. The "True Label" depicts the ground truth label: diagnostic (1) and non-diagnostic (0), whereas the "Pred Label" denotes the model prediction. A red arrow in the original image shows the location of the left atrium.
  • Figure 3: UMAPumap Visualization of embedding space representation of encoder of the three models. We report the 1st iteration's visualization among the $5$ iterations.