Automated neurological disorders detection is an essential step to ensure timely detection to benefit the patient and take necessary steps toward patient management system. Novel detection models for accurate disorder classification can remove burden from the health practitioners that looks promising from the resource management perspective.
Although an accurate automated model for neurological disorder detection is fundamental and researchers have contributed to the field of knowledge, there is no robust universal model that satisfies the practitioners and the policy makers altogether to enable real-world implication/ practice. Most models that have been developed fail to address the model's performance along with its explainability, interpretability and uncertainty matrices that can build trust and confidence for this imperative innovation/ automation. This PhD project might/aims to apply data driven deep learning techniques based on attention mechanism and Transformer (T) models, Grad-CAM (Gradient-weighted Class Activation Mapping), Local Interpretable Model-agnostic Explanations (LIME), Shapley Values and Monte Carlo dropout using electroencephalogram signals of the human brain. A review of research shows that the proposed methods are extremely powerful to find out the complicated features on the signal and successfully classify the disorders.
The outcome of this project might lead to accurate detection system of the neurological disorders with a universal trustworthy model. The techniques can also enable a commercial intervention.
For more information, please email the Graduate Research School or phone 0746 311088.