BEng, MEng, PhD
Personalised prescription based on individual-patient data using deep neural networks
My research interests and specialists are applications of machine learning and statistical analysis in medical and biological science, including personalised treatment for mental health care; biological image classification, segmentation, generation and clustering; biomedical signal processing; multi-criteria decision analysis (MCDA) in healthcare; smart sensors, wearable sensors, sensor integration and data fusion algorithms; NLP algorithms.
My recent research is focused on using machine learning, mainly deep neural networks to guide personalised treatment for individuals with dementia or depression. By applying machine learning and statistical models to both RCT and real-world longitudinal data (EHRs or clinical notes), we can predict post-treatment cognitive or depression scores, dropout risks, probabilities of experiencing side effects, etc., at an individual patient level. Furthermore, we can combine patient preference with treatment predictions for patient-centred decision-making using MCDA.
I have also been developing deep neural networks for cell morphological alteration detection. By analysing fluorescence images painted with various stains and bright-field images using both end-to-end deep learning models and extracted cell features, we can distinguish various treatment conditions in neuronal cell culture or co-culture of neurons and microglia. Given untreated neuronal images, we have developed several deep learning based generative models for artificial cell staining and generation.