Artificial Intelligence Techniques
The application of artificial intelligence, machine learning, and Big Data analytics is ubiquitous, touching all sectors of the society, and medicine and biomedical informatics are no different. While many extant AI techniques can be applied to clinical problems, there are many unique challenges associated with the medical domain that limit direct application. These include regulatory restriction such as HIPAA making it difficult to collect and share data, data persistence issues due to the sheer volume of data generated by the hospitals and other health care providers, and contending with missing or incomplete data. Moreover, there are epistemological issues surrounding medical conditions and diagnoses themselves, in that many common conditions have unclear definitions, making supervised learning challenging.
To address these issues, our lab is developing novel techniques in data analysis, machine learning, and time series analysis. These include creating new models that can handle uncertainty in data labels and data availability, mathematical techniques for handling missing or noisy data, and preserving the underlying structural or temporal nature of data within our analysis.