Prompt2Echo (P2Echo): Text-Prompted Multi-View Echocardiography Segmentation with One Model
Organizations: School of Biomedical Engineering, The University of British Columbia, Vancouver, BC V6T 1Z4, Canada · École polytechnique fédérale de Lausanne (EPFL), Lausanne, Switzerland · Faculty of Informatics and Data Science, University of Regensburg, 93053 Regensburg, Germany · an independent computer science researcher, Tehran, Iran ·
Abstract
Automated segmentation in multi-view echocardiography is essential for comprehensive cardiac assessment, yet remains challenging due to substantial anatomical and appearance variability across standard acoustic windows. In this paper, we present P2Echo, a unified framework capable of segmenting diverse cardiac structures across multiple standard echocardiographic views using a single model controlled by structured view-aware text prompts. Unlike recent foundation model adaptations that rely on implicit latent priors, P2Echo explicitly conditions the segmentation process via a novel Dynamic Image-Text Alignment (DyITA) module. DyITA integrates a differentiable routing mechanism and token differential operators to dynamically modulate decoder features according to the requested view and target structure. This allows the network to adaptively switch its processing pathways for different views without architectural changes. Extensive experiments on a large-scale internal dataset (seen during training) and held-out evaluation on three external datasets demonstrate that P2Echo matches or exceeds both view-specific specialists and state-of-the-art universal models at a fraction of their computational cost, offering an efficient, controllable solution for multi-plane cardiac analysis. Code is available at: https://github.com/moeinheidari7829/P2Echo.
Figures & tables
| View | Structures and descriptions |
| 4CH | LV : Bullet shaped, forms apex; LA : posterior to LV; RA : adjacent to LA, inferior to RV; RV : triangular, smaller than LV; MYO : inferoseptal, anterolateral, and apical segments. |
| 2CH | LV : base to apex; LA : superior to LV; MYO : anterior and inferior walls. |
| 3CH | LV : LV and LVOT in long axis; LA : visible in long axis; MYO : anteroseptal, inferolateral, and apical segments. |
| PSAX | LV : circular at mid ventricle; RV : crescent shaped around LV; MYO : six radial segments. |
| Dataset | 2CH | 3CH | 4CH | PSAX |
| Internal development | ||||
| Camus [ 3 ] | 10268 (1,2,3) | – | 10964 (1,2,3) | – |
| CardiacUDA [ 4 ] | – | – | 1834 (1,3,4,5) | – |
| EchoCP [ 15 ] | – | – | 548 (1,3,4,5) | – |
| EchoNet-Pediatric [ 16 ] | – | – | 6423 (1) | 8948 (1) |
| SegRWMA [ 17 ] | 941 (1) | 1053 (1) | 1116 (1) | – |
| Method | 2CH | 3CH | 4CH | PSAX | ||||||||
| mDice | mIoU | HD95 | mDice | mIoU | HD95 | mDice | mIoU | HD95 | mDice | mIoU | HD95 | |
| nnUNet [ 21 ] | {\color[rgb]{1,0,0}91.39}\pm 5.68 | |||||||||||
| SwinUNet [ 22 ] | ||||||||||||
| ResUNet [ 23 ] | ||||||||||||
| MissFormer [ 24 ] | ||||||||||||
| CENet [ 25 ] | {\color[rgb]{1,0,0}7.29}\pm 3.46 | |||||||||||
| Method | Params (M) | FLOPs (G) | HMC_QU | EchoNet-Dynamic | CardiacNet | |||
| mDice | HD95 | mDice | HD95 | mDice | HD95 | |||
| nnUNet [ 21 ] | 33.48 | 28.30 | {\color[rgb]{1,0,0}7.64}\pm 3.78 | |||||
| SwinUNet [ 22 ] | 27.17 | 6.16 | ||||||
| ResUNet [ 23 ] | 13.05 | 54.54 | ||||||
| MissFormer [ 24 ] | 42.46 | 9.89 | ||||||
| CENet [ 25 ] | 33.39 | 12.76 | {\color[rgb]{0,0,1}11.33}\pm 6.54 | |||||
| Ablation | Variant | mDice | HD95 |
| Text encoder | Qwen (0.6B) | 82.22 | 6.92 |
| Qwen (4B) | 82.29 | 6.95 | |
| DyITA components | w/o Dynamic Projection | 80.04 | 7.31 |
| w/o Dynamic Measure Kernel | 80.52 | 7.35 | |
| w/o Differential Operator | 80.03 | 7.84 | |
| Full DyITA | 82.22 | 6.92 |
| Condition | Example prompt | Dice (%) | IoU (%) |
| Canonical | “Segment the LV: The LV is fully visualized from base to apex.” | 93.53 6.83 | 88.39 8.84 |
| Clinical paraphrase | “Delineate the left ventricular blood pool.” | 80.48 23.88 | 72.62 26.76 |
| Concise | “Segment the left ventricle cavity.” | 35.83 36.23 | 28.58 30.95 |
| Ambiguous | “Segment the cardiac chamber.” | 0.01 0.23 | 0.01 0.12 |
| Irrelevant | “Segment the liver.” | 6.35 19.70 | 4.95 16.54 |
| Null | “No anatomical structure is requested.” | 0.00 0.00 | 0.00 0.00 |
| Method | Time (s/image) |
| nnUNet | 0.04057 0.00002 |
| SwinUNet | 0.00577 0.00002 |
| MissFormer | 0.01034 0.00001 |
| CENet | 0.02328 0.00010 |
| SAMUS | 0.06620 0.00024 |
| SAM3 (Zero-shot) | 0.19694 0.00004 |
Ultrasound Image Segmentation
Ultrasound image segmentation delineates organs, lesions, and other structures in ultrasound scans for measurement, diagnosis, and image-guided care.
See all 78 papers