Prostate Cancer Detection with mp-MRI
Multi-parameter magnetic resonance imaging (mp-MRI) data containing clinically significant prostate cancer is critical for automated PCa detection with high accuracy. This project focuses on designing novel Generative Adversarial Networks (GANs) for synthesizing high-quality mp-MRI images of clinically significant prostate cancer.
Automated Diagnosis in mp-MRI
Multi-parameter MRI is increasingly popular for prostate cancer detection and diagnosis, but interpreting unregistered 3D sequences such as ADC and T2w images remains time-consuming. This work introduces a series of recent deep convolutional neural networks for automated PCa detection and diagnosis.
Retinal Vessel Image Analysis
Retinal fundus images provide rich information for diagnosing eye diseases such as macular degeneration, diabetic retinopathy, and glaucoma. Accurate segmentation of retinal vessels remains challenging, especially for thin vessels, and better evaluation metrics are needed due to inter-observer variability.
Automated Renal Segmentation
We propose an automated renal segmentation method based on Maximally Stable Temporal Volume (MSTV). The approach segments the whole kidney, describes voxels with principal components, clusters them into cortex, medulla and pelvis, and refines each compartment to reduce noise.