ECHOPulse: ECG Controlled Echocardio-gram Video Generation
Authors: Yiwei Li, Sekeun Kim, Zihao Wu, Hanqi Jiang, Yi Pan, Pengfei Jin, Sifan Song, Yucheng Shi, Xiaowei Yu, Tianze Yang, Tianming Liu, Quanzheng Li, Xiang Li
ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Evaluations on three public and private datasets demonstrate state-of-the-art performance in ECHO video generation across both qualitative and quantitative measures. |
| Researcher Affiliation | Academia | 1Department of Radiology, Massachusetts General Hospital and Harvard Medical School 2School of Computing, University of Georgia 3Department of Computer Science and Engineering, University of Texas at Arlington |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Codes and examples can seen from https://github.com/levyisthebest/ECHOPulse_Prelease. |
| Open Datasets | Yes | Initially, we train the video tokenization model on the Webvim-10M dataset (Bain et al., 2021), followed by fine-tuning on the CAMUS dataset (Leclerc et al., 2019b) to adapt the tokenization for medical imaging tasks. For evaluation, we use the CAMUS (Leclerc et al., 2019a) and Echo Net-Dynamic (Ouyang et al., 2020) datasets. |
| Dataset Splits | Yes | A total of 473 videos were extracted from the dataset to serve as the test dataset, and all video generation evaluations were performed on this test dataset. Heartbeat explicitly specifies key aspects such as the video resolution, certain parameters, and the use of the CAMUS dataset with its defined train, test, and validation splits. |
| Hardware Specification | Yes | All experiments were conducted on 8 NVIDIA H-100 GPUs, each with 80GB of memory. |
| Software Dependencies | No | The paper mentions using T5X and Adam optimizer, but does not provide specific version numbers for these or any other software dependencies. |
| Experiment Setup | Yes | Table 4: Video tokenizer hyperparameters and Table 5: Video generation model hyperparameters (These tables contain specific hyperparameter values). |