<< Back to the abstract archive
Artificial Intelligence Generated High Quality Synthetic Data To Enhance Microsurgery Outcomes: A Proof-of-Concept Study
Berk B. Ozmen, Graham Schwarz
Cleveland Clinic Department of Plastic Surgery
2024-02-01
Presenter: Berk B. Ozmen
Affidavit:
This study is presented at ASRM 2024 Annual Meeting and the abstract is published at PRS-GO journal ASRM 2024 Annual Meeting abstracts section. After contacting Darlene, we were told to submit with this disclosure and that it would be the committee's decision to whether to accept or not to accept the submission. All work is original work of the authors.
Director Name: Graham Schwarz
Author Category: Fellow Plastic Surgery
Presentation Category: Clinical
Abstract Category: General Reconstruction
Background: The potential of artificial intelligence (AI) in improving surgical outcomes through predictive modeling is greatly hindered by the scarcity of large datasets, especially in specialized surgical subdisciplines such as microsurgery. Synthetic data, statistically designed to mimic real-world data with AI techniques, offers an innovative solution to overcome this challenge, enabling the generation of larger datasets from a limited number of cases.
Methods: We utilized Generative Adversarial Networks (GANs), an AI model using Python and TensorFlow to expand our single institute immediate lymphatic reconstruction dataset comprised of 133 cases with pre-operative, intra- operative and post-operative features. The dataset underwent comprehensive preprocessing, including normalization and one-hot encoding aimed to create a more suitable data for the model to operate effectively. A GAN was then trained on this data to generate a synthetic dataset that mimics the statistical properties of the original dataset.
Results: The GAN model was successful in generating a synthetic dataset of a total of 10,000 cases, significantly expanding the size of the original dataset while maintaining the statistical properties of the original dataset.
Conclusion: This synthetic dataset provides a powerful tool for AI-driven predictive modeling aimed at improving surgical outcomes. Our study underscores the potential of AI and GANs in generating synthetic surgical outcomes data in microsurgery. By significantly expanding the size of the original dataset, we have improved its utility for AI-driven predictive modeling, paving the way for advanced research aimed at improving surgical outcomes.
