Meng(Simon) Zhou (周蒙)

MSc. in Computer Science Candidate
Department of Computer Science
Intelligent Medical Image Computing Systems Lab (IMICS)
The Hospital for Sick Children (SickKids)
University of Toronto
Peter Gilgan Centre for Research and Learning (PGCRL) Building
686 Bay St, Toronto, Ontario, Canada

Email: simon DOT zhou AT mail.utoronto.ca

Google Scholar  /  GitHub  /  DBLP  /  ORCID
LinkedIn  /  Instagram  /  One-page CV  /  CV

- I am currently pursuing my MSc. degree in Computer Science at The University of Toronto .

- I am in the Intelligent Medical Image Computing Systems Lab, and I am very fortunate to be supervised by Dr. Farzad Khalvati.

- I am also a graduate research assistant at The Hospital for Sick Children, one of the top three paediatric health-care centres in the world. I am working in the Department of Neurosciences & Mental Health. My current research is focusing on Paediatric Low Grade Gliomas diagnosis.

- I obtained my Honours Bachelor's degree in Computing, Specialized in Computing and Mathematics from Queen's University.

- My undergraduate honours thesis "Domain Transfer Through Image-to-Image Translation in Prostate Cancer Detection" is supervised by Dr. Parvin Mousavi at the Medical Informatics Laboratory from Sept. 2021 to May 2022.

- I graduated from Yale Secondary School, a beautiful high school in British Columbia in June 2017.

- This is my main academic page, I have another website for personal, you may access from here.

In this paper, we have presented a novel approach for unpaired image-to-image translation of prostate mp-MRI for classifying clinically significant PCa, to be applied in data-constrained settings. First, we introduce domain transfer, a novel pipeline to translate unpaired 3.0T multi-parametric prostate MRIs to 1.5T, to increase the number of training data. Second, we estimate the uncertainty of our models through an evidential deep learning approach; and leverage the dataset filtering technique during the training process. Furthermore, we introduce a simple, yet efficient Evidential Focal Loss that incorporates the focal loss with evidential uncertainty to train our model. Experiments have shown the superior performance of the proposed approach.

We consider contextual bandits with a finite number of arms, where the contexts are independent and identically distributed d-dimensional random vectors, and the expected rewards are linear in both the arm parameters and contexts. We propose a truncated version of LinUCB and termed "Tr-LinUCB", which follows LinUCB up to a truncation time S and performs pure exploitation afterwards. The Tr-LinUCB algorithm is shown to achieve O(dlog(T)) regret if S=Cdlog(T) for a sufficiently large constant C, and a matching lower bound is established, which shows the rate optimality of Tr-LinUCB in both d and T under a low dimensional regime.

We propose an unpaired Image-to-Image translation framework to translate 3.0T prostate MRI data to 1.5T-like prostate MRI data, and then use a 3D Multi-Stream Convolutional Neural Network to classify the clinically significant Prostate Cancer.

We introduce a novel Dilated Residual Attention Network for the medical image fusion task. Our network is capable to extract multi-scale deep semantic features. Furthermore, we propose a novel fixed fusion strategy termed Softmax-based weighted strategy based on the Softmax weights and matrix nuclear norm. Extensive experiments show our proposed network and fusion strategy exceed the state-of-the-art performance compared with reference image fusion methods

We develop a new optimization framework, genetic-based evolutionary strategy, to optimize the convolutional neural networks architecture for the Chest X-Ray classification task. Experiments show that the proposed framework overperforms the results from the pre-trained model on ImageNet and fine tune on the X-Ray images.

We compare the performance of various deep models, ranging from traditional machine learning methods, CNN-based deep learning methods to the modern state-of-the-art Vision Transformer (ViTs) models. The results have showen that ViTs achieved the best performance in X-Ray classification task, and transfer learning improved ViT by a large margin.

Pitch-and-Poster Presentation at the 2022 Imaging Network Ontario Symposium.

Poster Presentation at Vector Institute Research Symposium. We take the 3.0T MRI images from the “ProstateX” challenge; translate to 1.5T-like MRI images based on the Cycle-GAN framework; and train a 3D Convolutional Neural Network for Prostate Cancer classification on translated images for the local use.