- 64 views
- 114 downloads
Comparative analysis of data synthesis algorithms for liver cirrhosis classification
-
- Author(s) / Creator(s)
-
Liver cirrhosis is a serious global health issue, causing a significant number of fatalities each year. Liver biopsy, the gold standard for diagnosis and staging, is an invasive procedure with potential complications and sampling errors. Noninvasive methods are being explored, but they are in early stages or require more resources. Artificial Intelligence (AI) has shown promise in healthcare, contingent on well-curated medical data. However, real-world medical data is often limited, leading to the use of synthetic data for training AI algorithms. Synthetic data generation (SDG) offers a privacy-preserving approach, preserving information while not containing original data. This study evaluates various SDG algorithms, including statistical methods and variations of Generative Adversarial Network (GAN), on Liver Function Tests (LFT) datasets. The aim is to augment the available dataset, providing diverse samples for training Machine Learning (ML) models through synthetic data based on actual samples. This approach improves the model’s understanding of underlying patterns and characteristics, resulting in more accurate liver cirrhosis diagnosis. The study emphasizes the importance of using laboratory test results for liver cirrhosis diagnosis as they offer a cost-effective alternative to invasive procedures. In this work, three tabular data generation algorithms are used namely CTGAN, Gaussian Copula, and CopulaGAN. Based on certain quantitative and qualitative methods, I present an analysis and evaluation of the most prominent algorithm for tabular data generation for Liver Cirrhosis Classification. Leveraging synthetic data to refine AI models, this research aims to contribute to advancements in liver disease diagnosis and treatment.
-
- Date created
- 2023
-
- Type of Item
- Research Material