4 Parallel AI Healthcare Workshop

    2-5pm, 5th July 2018, Innovation 4.0, NUS

  • Workshop 1:

    Secondary usage of Critical Care EMR data for care improvement (Max 30 pax)

    Venue: AI Singapore, level 1

    Prof Leo Celi (MIT) & Data Scientist team from MIT

    The 3-hour workshop will introduce the participants to the Medical Information Mart for Intensive Care or MIMIC database. MIMIC is a collaborative project of the Laboratory for Computational Physiology at the Massachusetts Institute of Technology, and the Beth Israel Deaconess Medical Center, a teaching hospital of Harvard Medical School. It contains de-identified data collected from patients admitted to the intensive care units of BIDMC including vital signs, physiologic waveforms, laboratory and radiology reports, time-stamped medications and clinical notes, and soon, medical images linked to the rest of the health records. With close to 10,000 users from academia, industry, and government around the globe, it represents a pioneering venture that provides a critical source of reliable data for clinical research, education and innovations in the medical device and decision support tools. During the workshop, the data elements will be explored through visualization and examples of prediction models will be presented.


    Note: Participants are advised to bring along your own laptop for the best experiences of the workshop. For any inquiries, please contact Wei-Hung Weng from MIT at ckbjimmy@mit.edu.

    Workshop 2 (Comprises 2 sessions):

    Part 1: Application of AI Technologies in Medical Images (Max 30 pax)

    Venue: AI Singapore, level 1

    Asst Prof Mengling 'Mornin' Feng (NUS)

    Application of AI Technologies in Medical Images (Part 1)

    Artificial intelligence (AI) is the buzzword for all things relating to technology these days. In particular, healthcare is seen as an area in which AI may be gainfully deployed to improve medical care, especially with big data, exponential computing power and a burgeoning demand on healthcare systems due to aging populations. In this workshop, I will share how deep learning based AI technologies can be applied to analyze medical images, in particular, the mammogram images. My team ranked top 5 in the DREAM digital Mammogram challenged funded by the US white house fund and FDA.


    I will share the fundamentals on how deep neural networks, in particular the Convolution Neural Networks (CNN). We will go through some hands-on exercise to gain a quick understand how CNN works and how the architecture of CNN can be built. Then, I will share the lessons learned via our mammogram project. Only pen and paper will be required.


    Note: The first part of the workshop is designed to be an introductory course to CNN and deep learning. It is meant for physicians and scientists, who will like to learn about CNN. Minimum technical foundation is required. The workshop is not suitable for data scientists, who are already familiar with CNN.

    For any inquiry, you may reach Dr. Feng at ephfm@nus.edu.sg.


    Workshop 2 (Comprises 2 sessions):​

    Part 2: Image Classification with Digits (Max 30 pax)

    Ritchie Ng, Head of Deep Learning, Ensemble Capital (NVIDIA)

    Image Classification with Digits (Part 2)

    The NVIDIA Deep Learning Institute (DLI) trains developers, data scientists, and researchers on how to use artificial intelligence to solve real-world problems across a wide range of domains. In the deep learning courses, you’ll learn how to train, optimize, and deploy neural networks using the latest tools, frameworks, and techniques for deep learning.


    In “Image Classification with DIGITS” mini-course, you will learn how to train a deep neural network to recognize handwritten digits by loading image data into a training environment, choosing and training a network, testing with new data, and iterating to improve performance.


    Note: For the second part of the workshop, participants are advised to bring along your own laptop for the best experiences of the workshop.

    Workshop 3:

    Sequencing Analysis Workflows for Precision Medicine and Data-intensive Genomics (Max 30 pax)

    Venue of the workshop has been changed to NUHS Tower Block, T08-03

    Dr Jason Pitt (CSI, NUS)

    For decades, researchers and clinicians have envisioned genetic information playing a inevitable — and significant — role in medical practices. Decreasing sequencing costs and continuous technological improvements now allow us rapidly characterize individuals’ genetic variation, bringing precision medicine closer to actualization. However, the subsequent influx of data increases the cost and complexity of downstream analyses. Substantial hardware resources (CPU, memory, disk space, etc.) are necessary but not sufficient to leverage these data streams. Sophisticated software solutions are required to harness the scientific and clinical opportunities enabled by contemporary sequencing technologies — particularly at scale. To this end, we have developed SwiftSeq (https://swiftseq.uchicago.edu/), a flexible, portable, and intuitive workflow infrastructure for the analysis of DNA sequencing data. By design, this approach allows even those with limited programmatic experience to conduct small- and large-scale genomic analyses.


    By participating in this workshop, attendees will better understand:

    1. The computational challenges of genetic variant identification.
    2. How to design and run variant calling workflows using SwiftSeq.
    3. How to use SwiftSeq to effectively scale large analyses.
    Note: Participants are advised to bring along your own laptop for the best experiences of the workshop. For any inquiries, please contact Dr. Jason Pitt at csijjp@nus.edu.sg.

    Workshop 4:

    Deploying AI Solutions in Real Clinical Practices (Max 30 pax)

    Venue of the workshop has been changed to NUHS Tower Block, T08-04

    Dr Ngiam Kee Yuan (CTO, NUHS)

    Healthcare data is notorious for being haphazard, messy and often incomplete. To use this data effectively, it requires extensive pre-processing to obtain reliable results that are representative of true patient characteristics. This is the most important and time-consuming step in machine model development. In particular, when training deep learning models to be deployed in healthcare, it is especially important that these tools are applied to 'clean' data, and curated to correctly represent potential mechanistic relationships between patient features. This workshop will cover key aspects of data validation and curation, in the context of training deep learning models.


    Note: Participants are advised to bring along your own laptop for the best experiences of the workshop. For any inquiries, please contact Dr. Ngiam at kee_yuan_ngiam@nuhs.edu.sg.