Exploring Feature Enhancement for Lesion Detection and Segmentation

The aim of this PhD thesis is to explore artificial intelligence models that more accurately detect and segment lesions of varying sizes in medical images by leveraging feature enhancement techniques within advanced deep neural networks. This thesis included five published works in methodological chapters proposing novel deep-learning models that incorporate various feature enhancement techniques for monoplane/multiplane brain tumour detection, blood cell detection, liver tumour segmentation, and cancer cell instance segmentation, demonstrating higher precision in lesion detection and/or segmentation while equivalent computation to general-purpose object detection and/or segmentation models, and making valuable contributions to extending existing knowledge in clinical computer-aided diagnosis.