IEEE
IEEE HomeSearch IEEEShopWeb AccountContact IEEE
MembershipPublicationsServicesStandardsConferencesCareers/Jobs
 

Monthly Meetings

Officers

IEEE Women in Engineering

Student Activities

Annual Awards Banquet

Congressional Visits Day


Southern New Jersey IEEE


spacer spacer

The Southern Jersey Professional Societies
American Institute of Aeronautics and Astronautics (AIAA)
Institute of Electrical and Electronics Engineers (IEEE),
International Council of Systems Engineers (INCOSE),
International Test and Evaluation Association (ITEA), National Society of Black Engineers (NSBE), and the Human Factors Ergonomics Society (HFES)

Wednesday Sept 20, 2017

Presents

Deep Learning vs. Bayesian Thinking Application to Image Segmentation

Speaker: Dr. Nidhal Bouaynaya, Dept. of Electrical and Computer Engineering, Rowan University

Biomedical image analysis and segmentation is challenging because of the unpredictable appearance and shape of tumors from multi-modal imaging data. In this talk, we investigate two approaches: Mathematical modeling and optimization vs. deep learning; more specifically, partial differential equations vs. convolutional neural networks. Mathematically, we propose a deformable model based on the variational level set method (LSM) and non-negative matrix factorization (NMF). We describe the use of NMF, an algorithm based on decomposition by parts that can reduce the dimension of images from thousands of pixels to a handful of regions. Coupled with the variational framework, NMF-LSM is a powerful image segmentation method with pixel-level accuracy. In the second approach, we propose an automatic segmentation method based on Convolutional Neural Networks (CNNs) with a deep architecture. The two approaches are validated on the Multimodel Brain Tumor Segmentation Challenge (BRATS) 2016 database. We use the Dice Similarly coefficient and Haussdorff measure to assess the three tumor regions: whole tumor, core tumor and necrosis. NMF-LSM showed superior performance for whole tumor and necrosis in Hausdorff measure, whereas CNN showed superior performance for the tumor core in Dice metric. In our ongoing efforts, we are integrating both approaches towards a complete clinic-ready suite of MR analysis and display tools, named coordinates and volumetrics for brain tumors.

    Location: Blue Heron Pines Golf Club
    550 Country Club Dr, Egg Harbor City, NJ 08215

    (609) 965-1800

    Directions: www.blueheronpines.com

    TIME:  

    6:00 PM    Social Time
    6:30 PM    Dinner
    7:45 PM    Speaker, approx.

    MENU: Buffet style dinner

    COST:

    $19 AIAA/IEEE/INCOSE Members
    $23 ITEA/NSBE/HFES/Guests
    $15 Students

    Reservations due by noon Tuesday, June 27

    Joseph Burns 609-338-8789  (C) jfburns54@faa.gov
    Mike Konyak 609-280-8179 (C) snj.aiaa@gmail.com
         
     


    Keep in touch at the SNJ virtual community http://www.ieeecommunities.org/snj

    (IEEE web account needed for access)


 
Webmaster for SNJ IEEE: (olanm@computer.org)
(Modified: 11-Sept-2017)