Scientific Computing Seminar

Date and Place: Thursdays and hybrid (live in 32-349/online via Zoom). For detailed dates see below!

Content

In the Scientific Computing Seminar we host talks of guests and members of the SciComp team as well as students of mathematics, computer science and engineering. Everybody interested in the topics is welcome.

List of Talks

Event Information:

  • Thu
    18
    Feb
    2021

    SC Seminar: Mitesh Mittal

    11:30Online

    Mitesh Mittal, TU Kaiserslautern

    Title: Deep learning approaches for medical image data segmentation and
    classification

    Abstract:

    Deep neural networks are now the state-of-the-art machine learning
    models across a variety of areas, from computer vision, recommendation
    systems, natural language processing, and are also widely deployed in
    academia and industry. Their potential for use in medical imaging
    technology, medical data analysis, medical diagnostics and healthcare
    in general, are slowly being realized. We will look at recent advances
    and some associated challenges in deep learning applied to medical
    image segmentation and image classification.

    The image classification problems arise in diabetic retinopathy
    diagnosis. Methods based on UCI experiments often fail on real diabetic
    retinopathy data. Bayesian deep learning enables the estimation of
    model uncertainty and sets benchmarks based on the uncertainty in
    classification. The uncertainty level determines whether an expert
    referral is necessary for better diagnosis. The BDL method outperforms
    current UCI experiments which overfit their uncertainty to the dataset
    and provide better diagnosis diabetic retinopathy diagnosis.

    In medical image segmentation, we look at deep Convolutional Neural
    Networks (CNN) that use data augmentation on available annotated
    samples to give better performance. This model can be trained end-to-
    end from very few images and the architecture takes into account the
    context and localization of the image patches.
    This approach outperforms the prior best method (a sliding-window
    convolutional network) on the ISBI challenge for segmentation of
    neuronal structures in electron microscopic stacks by large margins.

    The striking performance of these methods motivates their application
    to other complex problems in medical image data processing.

    How to join:

    The talk is held online via Jitsi. You can join with the link https://jitsi.uni-kl.de/SciCompSeminar_13. Please follow the rules below:

    • Use a chrome based browser (One member with a different browser can crash the whole meeting).
    • Mute your microphone and disable your camera.
    • If you have a question, raise your hand.

    More information is available at https://www.rhrk.uni-kl.de/dienstleistungen/netz-telefonie/konferenzdienste/jitsi/.