The application of AI/ML techniques in healthcare data has gained remarkable popularity in recent times. However, a significant challenge lies in the scarcity of data due to the stringent limitations associated with collecting patient information. This process is governed by various constraints, including privacy and ethical considerations. Due to data scarcity, the reported accuracies of AI/ML models can be notably influenced by the selection and availability of data, potentially introducing biases. Additionally, comparing different models based solely on reported accuracy becomes challenging, as these numbers may not comprehensively represent a model's intrinsic abilities. This limitation complicates decision-making processes that rely solely on these metrics.

The objective of this project is to create a statistical framework capable of estimating the inherent accuracy of AI/ML models. Additionally, this project seeks to explore the disadvantages of commonly employed methods to address data scarcity, including k-fold cross-validation and data augmentation.


Electrical Engineering and Telecommunications

Research Area

Statistical machine learning

The project will be conducted within the Signals, Information and Machine Intelligence (SIMI) Lab, situated within the School of Electrical Engineering and Telecommunications. This dynamic group comprises two full-time academic staff members, a Postdoctoral Research Fellow, and multiple research students. Collectively, they are immersed in diverse research endeavors that involve the application of AI/ML techniques to an array of disciplines, encompassing audio processing, image processing, and time-series data analytics.

Upon the completion of this project, students will acquire a comprehensive knowledge of AI/Machine learning (ML) techniques within healthcare applications. Furthermore, they will substantially enhance their proficiency in technical research methods and programming (Python/Julia), particularly in utilizing AI/ML frameworks (Tensorflow/ Pythorch/Flux). By the project's conclusion, students are expected to have successfully developed statistical methods and validated previously published AI/ML algorithms.