Wednesday, July 21, 2021 - 11:30 to 13:00
SCHOOL OF COMPUTER SCIENCE
The School of Computer Science at the University of Windsor is pleased to present …
PhD. Seminar Presentation by: Yahya Alzahrani
Date: Wednesday July 21 ,2021
Time: 11:30 am-1:00 pm
Meeting URL: https://zoom.us/j/93224615822?from=addon
Passcode: If interested in attending this event, contact the Graduate Secretary at email@example.com
Modern Computer-Aided Diagnosis (CAD) systems have brought huge potential for an early diagnosis of breast tumors which are the women’s leading causes of death worldwide. Utilizing the available Breast Ultrasound images (BUS), deep learning algorithms have shown a vast success in segmentation and classification tasks. yet, this is a challenging task since Ultrasound(US) images are very noisy with class imbalanced data distribution and intensity inhomogeneity. To address these issues of BUS images, we proposed a new variant of the U-Net architecture with deep encoder. We also increased the depth of the network by adapting the Residual Blocks (RBs). Our proposed approach includes pre-processing stage, feature extraction based residual blocks, convolution-concatenation path and simple decoder to reconstruction the extracted features. In this paper, we utilized two publicly available BUS datasets: BUSIS and UDIAT. To resolve the issue of the vanishing gradient while down sampling the features, we expanded the network’s width by adapting convolution path instead of the concatenation path in the original U-net. Our model has shown a better performance as compared to the plain U-Net and other models like selective Kernel-U-Net and Efficient-U-Net. The proposed model achieved Dice coefficient and IOU of 91.5 and 84.6 on BUSIS, and 0.81 and 0.69 on UDIAT, respectively.
Keywords: BUS Images, Image Segmentation, Neural Networks, Residual U-Net.
External Reader: Dr. Esam Abdel-Raheem
Internal Reader: Dr. Christie Ezeife
Internal Reader: Dr. Dan Wu
Advisor: Dr. Boubakeur Boufama
PhD Seminar Announcement
5113 Lambton Tower 401 Sunset Ave. Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716 firstname.lastname@example.org (working remotely)