From Sparse to Precise: A Practical Editing Approach for Intracardiac Echocardiography Segmentation

Accurate and safe catheter ablation procedures for atrial fibrillation require precise segmentation of cardiac structures in Intracardiac Echocardiography (ICE) imaging. Prior studies have suggested methods that employ 3D geometry information from the ICE transducer to create a sparse ICE volume by placing 2D frames in a 3D grid, enabling the training of 3D segmentation models. However, the resulting 3D masks from these models can be inaccurate and may lead to serious clinical complications due to the sparse sampling in ICE data, frames misalignment, and cardiac motion. To address this issue, we propose an interactive editing framework that allows users to edit segmentation output by drawing scribbles on a 2D frame. The user interaction is mapped to the 3D grid and utilized to execute an editing step that modifies the segmentation in the vicinity of the interaction while preserving the previous segmentation away from the interaction. Furthermore, our framework accommodates multiple edits to the segmentation output in a sequential manner without compromising previous edits. This paper presents a novel loss function and a novel evaluation metric specifically designed for editing. Cross-validation and testing results indicate that, in terms of segmentation quality and following user input, our proposed loss function outperforms standard losses and training strategies. We demonstrate quantitatively and qualitatively that subsequent edits do not compromise previous edits when using our method, as opposed to standard segmentation losses. Our approach improves segmentation accuracy while avoiding undesired changes away from user interactions and without compromising the quality of previously edited regions, leading to better patient outcomes.

1. 1.

   Annotations typically take the form of contours instead of masks, as the structures being segmented appear with open boundaries in the frames.

2. 2.

   Contours are inferred from the predicted mask \(\hat{y}\).

3. 3.

   We do not utilize the CAS contours during training and only use them for testing because the CAS contours do not align with the segmentation meshes y.

Authors and Affiliations

1. Siemens Healthineers, New Jersey, USA

   Ahmed H. Shahin, Yan Zhuang & Noha El-Zehiry

2. University College London, London, UK

   Ahmed H. Shahin

3. National Institutes of Health Clinical Center, Maryland, USA

   Yan Zhuang

4. Wipro, New Jersey, USA

   Noha El-Zehiry

Corresponding author

Correspondence to Ahmed H. Shahin .

Editors and Affiliations

1. Icahn School of Medicine, Mount Sinai, NYC, NY, USA, Tel Aviv University, Tel Aviv, Israel

   Hayit Greenspan

2. Emory University, Atlanta, GA, USA

   Anant Madabhushi

3. Queen's University, Kingston, ON, Canada

   Parvin Mousavi

4. The University of British Columbia, Vancouver, BC, Canada

   Septimiu Salcudean

5. Yale University, New Haven, CT, USA

   James Duncan

6. IBM Research, San Jose, CA, USA

   Tanveer Syeda-Mahmood

7. Johns Hopkins University, Baltimore, MD, USA

   Russell Taylor

Work done while the authors were employed by Siemens Healthineers.

Disclaimer: The concepts and information presented in this paper are based on research results that are not commercially available. Future commercial availability cannot be guaranteed.

Reprints and permissions

© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG

Policies and ethics