EchoONE: Segmenting Multiple echocardiography Planes in One Model

2024-12-04

arXiv.org

AI Summary

Echocardiographic multi-plane segmentation (MPS) faces significant challenges due to substantial anatomical variations across views and the prevailing reliance on view-specific AI models, resulting in a lack of generalizable, efficient solutions. To address this, we propose the first end-to-end,通用 (unified) MPS framework based on the Segment Anything Model (SAM) architecture. Our method introduces two key innovations: (1) a semantic-aware Prior-Composable Mask (PC-Mask) module for dense, anatomy-informed prompt generation; and (2) a Local Feature Fusion and Adaptive CNN (LFFA) branch to enhance cross-view feature consistency and robustness. The single unified model achieves accurate, simultaneous segmentation across standard views—including apical four-chamber (A4C), apical two-chamber (A2C), and parasternal short-axis (PSAX). Evaluated on multiple internal and external multicenter datasets, it consistently achieves state-of-the-art performance, demonstrates strong robustness across imaging devices and acquisition protocols, and significantly reduces deployment complexity—establishing a scalable foundation for automated cardiac structural analysis in clinical practice.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In clinical practice of echocardiography examinations, multiple planes containing the heart structures of different view are usually required in screening, diagnosis and treatment of cardiac disease. AI models for echocardiography have to be tailored for each specific plane due to the dramatic structure differences, thus resulting in repetition development and extra complexity. Effective solution for such a multi-plane segmentation (MPS) problem is highly demanded for medical images, yet has not been well investigated. In this paper, we propose a novel solution, EchoONE, for this problem with a SAM-based segmentation architecture, a prior-composable mask learning (PC-Mask) module for semantic-aware dense prompt generation, and a learnable CNN-branch with a simple yet effective local feature fusion and adaption (LFFA) module for SAM adapting. We extensively evaluated our method on multiple internal and external echocardiography datasets, and achieved consistently state-of-the-art performance for multi-source datasets with different heart planes. This is the first time that the MPS problem is solved in one model for echocardiography data. The code will be available at https://github.com/a2502503/EchoONE.

Problem

• Segmentation of multiple echocardiography planes in one model.
• Reducing repetition and complexity in AI model development.
• Achieving state-of-the-art performance across diverse heart planes.

Innovation

• SAM-based segmentation architecture for echocardiography
• PC-Mask module for semantic-aware dense prompts
• CNN-branch with LFFA module for SAM adaptation