PhD Doctoral Proposal Presentation Announcement by Abdala Nour:"Deep learning approach for Mammography images of Breast Cancer segmentation and classification hybrid with Active contour and image processing techniques"

 

Date: Thursday, February 23, 2023

Time:  12:00 pm - 1:00 pm

Location: Essex Hall, Room 105

 

Reminders: 1. Two-part attendance mandatory (sign-in sheet, QR Code)

2. Arrive 5-10 minutes prior to the event starting - LATECOMERS WILL NOT BE ADMITTED. Due to demand, if the room has reached capacity, even if you are "early" admission is not guaranteed.

3. Be respectful of the presenter by NOT knocking on the door for admittance once the door has been closed whether the presentation has begun or not (If the room is at capacity, overflow is not permitted (ie. sitting on floors) as this is a violation of the Fire Safety code).

4. Be respectful of the decision of the advisor/host of the event if you are not given admittance. Remember, The School of Computer Science has numerous events occurring in the near future.

Breast cancer is the most common aggressive cancer in women because of uncontrolled cell growth in the body. While breast cancer has declined over time, it remains women's second leading cause of cancer death. However, early detection of this cancer can reduce its aggressiveness. But it is challenging to identify breast masses due to the low contrast, blurry boundaries, and large amount of noise and artifacts in breast mammogram images. This leads to misguided segmentation of breast mass and tissues. Therefore, it remains a challenging task to reduce mortality rates by automatically and accurately segmenting masses. In order to improve the efficiency and accuracy of understanding medical images, various algorithms have been developed to get fully automatic segmentations from medical images. These include utilizing deep neural network (DNN) techniques such as convolutional neural networks (CNNs), which are updated later to identify regions of interest (ROIs) in mammography images. Thus, deep convolutional neural networks (DCNNs) have proved successful in the field of semantic segmentation and classification of medical images. However, DNN's predictions can be quite rough in some cases due to artifacts and blurry noises. The active contour model (ACM) is used in a variety of clinical imaging fields to draw smooth shapes in the image and to form closed contours for the area. Using a fully convolutional neural network (FCN) hybrid with an active contour model, we can combine a region-based interest function and a phase-based energy function to segment breast cancer images more accurately and reliably. In this paper, we propose a robust multiobjective optimization technique to segment breast masses from mammographic images. Specifically, mammography images can be analyzed with deep learning methods to extract shape, texture, and color information. These features can then be used as input to an active contour model, which can be used to segment breast mass regions from mammography images. The mammographic image datasets are collected from benchmark datasets like the Digital Database for Screening Mammography (DDSM). Furthermore, in order to evaluate our method, we developed different types of deep convolutional neural network models designed for medical image segmentation, namely FCN, U-Net, VGG-16, and Resnet50, to measure the performance of each model.

 

Keywords: Semantic Segmentation, Convolutional Neural Network(CNN), Fully Convolutional Neural Network (FCN), U-Net architecture, Active Contour Model (ACM), Regions Of Interest (ROIs).

 

Internal Reader: Dr. Dima Alhadidi

Internal Reader: Dr. Imran Ahmad

External Reader: Dr. Mohamed Belalia        

Advisor:  Dr. Boubaker Boufama