- Alfonso Pineda Cedillo | A01660394
- Mariana Ivette Rincón Flores | A01654973
- Salvador Mendoza Tinoco | A01067783
- Karla González Sánchez | A01541526
- Álvaro Morán Errejón | A01638034
This project focuses on the segmentation of the left ventricle of the heart, employing two distinct approaches: the use of masks and landmark identification. The implementation leverages convolutional neural networks U-NET, trained on the "EchoNet-Dynamic" dataset. This specific dataset contains echocardiogram videos, providing a robust foundation for network training.
The primary purpose of this initiative is to evaluate and compare the effectiveness of the two segmentation methods. We aim to determine which yields superior results, with potential applications in the medical domain. The choice between mask-based segmentation and landmark identification will be informed by quantifiable and validated outcomes derived from real-world data training, contributing to informed decision-making in future medical developments related to left ventricle heart segmentation.
Throughout this project, we have prioritized adherence to industry laws, regulations, and ethical principles inherent to the challenge context. To ensure this, we have meticulously followed the following aspects:
We have ensured that all our actions related to data acquisition, storage, and processing are carried out in strict accordance with relevant regulations. This process ensures comprehensive data analysis, facilitating information extraction and the identification of significant patterns. Additionally, we have observed and respected copyright and intellectual property rights by legally and ethically utilizing the "EchoNet-Dynamic" dataset. For greater transparency, we have thoroughly documented the methodologies and approaches used, enabling others to understand and assess our decisions in this domain.
We have conducted a thorough review and rigorously adhered to all the specific regulations and norms established by the presented project. This approach ensures that our solution complies with each of the rules and requirements outlined by the challenge. By strictly adhering to these regulations, we aim not only to achieve the project's objectives but also to demonstrate our commitment to integrity and legality at every stage of development.
This document reflects our unwavering commitment to legal and ethical integrity in every phase of this project, ensuring transparency and accountability in all our actions.
- EchoNet-Dynamic dataset.
- Notebook containing the challenge solution using the mask methodology.
- Notebook containing the challenge solution using the landmark methodology.
- Exported U-Net model with
.h5extension. - Image folder (frames, masks, contours, etc.) used for model training.
- Image folder showcasing segmentation results from the model.
- Notebooks containing auxiliary functions for the challenge solution.
- Project results (final report).