Warning! OUTDATED BROWSER DETECTED!   Please update your browser immediately for a better experience on this website. Learn More
21/xsl/MobileMenu.xsltmobileNave880e1541/WorkArea//http://www.rsna.org/RSNANewsDetailWireframe.aspx?pageid=15319&id=22957&ekfxmen_noscript=1&ekfxmensel=falsefalsetruetruetruefalsefalse10-18.0.0.0730truefalse
  •  
     
  • Machine Learning Plays Central Role at RSNA 2017

    By Mike Bassett


    November 1, 2017

    Machine Learning (ML) and the role it will play in the future of radiology will be central to a broad scope of programming at RSNA 2017.

    Along with numerous educational and scientific sessions exploring the revolutionary technology from every angle, RSNA is featuring a Machine Learning Community in the Learning Center and a Machine Learning Showcase in the Technical Exhibits hall.

    Despite the technology’s growing presence in healthcare, there are still obstacles to overcome before ML is fully embraced by radiology, said Curtis Langlotz, MD, the RSNA Board of Directors liaison for information technology and annual meeting.

    “No question — machine learning will change the way radiologists practice in the years ahead, sometimes dramatically,” said Dr. Langlotz, a professor of radiology and biomedical informatics and associate chair for Information Systems in the Department of Radiology at Stanford University. “But there is much work to be done before ML becomes commonplace.”

    Along with facing regulatory issues, ML requires large, labeled image data sets for big data processing. But while most radiology practices have millions of imaging studies, most are not labeled, he said.

    “There is a major research focus right now on how to automate, or at least partially automate, the image labeling process using information from the radiology report or the medical record,” Dr. Langlotz said. “We also need a trained labor force to build these algorithms.”

    While hundreds of graduate students at Stanford are learning such techniques, this type of training needs to be developed across the specialty, he said.

    “Truly mastering the problem solving requires mentorship from those who have done it before at a high level,” Dr. Langlotz said. “We have some great resources in Silicon Valley, but that’s not yet true across the board.”

    Machine Learning Community

    To that end, RSNA 2017 is featuring a Machine Learning Community with many ML educational opportunities.

    The RSNA Deep Learning Classroom presented by NVIDIA DLI will give attendees a rang of hands-on courses to engage with ML tools, write algorithms and improve their understanding of ML technology.

    A National Cancer Institute (NCI)-sponsored exhibit, “Crowds Cure Cancer: Help Annotate Data from The Cancer Imaging Archive” offers RSNA attendees the opportunity to use annotation tools to label data sets in NCI’s cancer imaging archive for use in ML research.

    The Machine Learning Community will also include ML Hardcopy Backboard posters, select posters from Research & Education (R&E) Foundation grant recipients and authors of ML articles published in Radiology as well as posters from the organizing committee and winners of the RSNA Pediatric Bone Age Challenge.

    Visit the Machine Learning Community in the Learning Center, Lakeside Center East, Level 3. Exhibits are open throughout the week.

    Machine Learning Sessions at RSNA 2017

    Visit Meeting.RSNA.org for a complete schedule of sessions.

    • Deep Learning & Machine Intelligence in Radiology — RC153

    • Hot Topic Session: Machine Learning and Artificial Intelligence in Lung Imaging — SPSH20

    • Introduction to Machine Learning and Texture Analysis for Lesion Characterization (Hands-on) — RCA25

    • Neuroradiology (Machine Learning and Deep Learning) — SSJ19

    • Platforms and Infrastructures for Accelerated Discoveries in Machine Learning and Radiomics — RCC42

    • Leveraging Machine Learning Techniques and Predictive Analytics for Knowledge Discovery in Radiology (Hands-on) — RCA53

    • Deep Learning — An Imaging Roadmap — RCC45

    • Machine Learning Techniques for Automated Accurate Organ Segmentation and Their Applications to Diagnosis Assistance — ML001-EX-B; Hardcopy Backboard

    • What Was Changed in Machine Learning in Medical Image Analysis After the Introduction of Deep Learning? — ML-106-ED-X; Digital Education Exhibit

    • RSNA Deep Learning Classroom, presented by NVIDIA DLI — ML001-EX-B

    Machine Learning Showcase

    A focal point of the Technical Exhibits, North Exhibit Hall, the Machine Learning (ML) Showcase gives attendees an opportunity to learn about the latest ML technology and network with companies on the forefront of ML advancements.

    The showcase will feature a Machine Learning Theater, offering ML presentations daily between 11 a.m. and 2 p.m. Search the RSNA Meeting Program for a full lineup of theater presentations.

    The Machine Learning Showcase is sponsored by Carestream Health, Google Cloud and Zebra Medical Vision.

    An Expert Defines the Terms

    Because of the rapid growth of the technology, there may be confusion over the terms machine learning (ML), Artificial Intelligence (AI) and deep learning, said Paul E. Kinahan, PhD, who will moderate the 2017 RSNA/American Association of Physicists in Medicine (AAPM) Symposium.

    “Artificial Intelligence and machine learning are not exactly the same thing, but there is a large overlap,” Dr. Kinahan said.

    AI is essentially the simulation of intelligent behavior in computers while ML refers to the algorithm or method used in AI, he said. “ML algorithms can be used for many different tasks, one of which is supporting AI. AI, however, can use methods other than ML,” Dr. Kinahan said. “Further confusing the issue is the emergence of deep learning which is like the Godzilla of machine learning — much larger and slower to get going, but much more effective it seems.” Deep learning is an area of ML research with the objective of moving ML closer to AI, he added.

    Accurate Algorithms Focus of RSNA Machine Learning Challenge

    New this year, RSNA introduced a Machine Learning (ML) Challenge designed to showcase methods for creating algorithms to address clinical problems in radiology, in this case automating the assessment of pediatric bone age based on hand radiographs.

    The RSNA Pediatric Bone Age Challenge is an online competition that was held from August through October. Three teams submitting the most accurate algorithms will receive awards Monday, Nov. 27, in the Machine Learning Showcase in the Technical Exhibits Hall.

    Launched by the RSNA Radiology Informatics Committee, the competition used skeletal data sets from Stanford Children’s Hospital, Colorado Children’s Hospital and the University of California, Los Angeles. Participants were given a training set of hand radiographs and corresponding skeletal ages. The more than 250 participants who entered the challenge worked in 29 teams to submit algorithms.

    The challenge was hosted on the MedICI platform for medical image computer challenges built by Jayashree Kalpathy-Cramer, PhD, an assistant professor of radiology at Harvard Medical School and Massachusetts General Hospital, and colleagues.

    “Challenges can be an effective means to comprehensively assess the performance of algorithms by comparing them on common, sufficiently large and diverse datasets using realistic tasks and valid evaluation metrics,” said Dr. Kalpathy-Cramer, who served on the ML Challenge Organizing Committee.

    Submissions to the ML Challenge will be exhibited in the Machine Learning Showcase and the Machine Learning Community at RSNA 2017.




    Langlotz
    Langlotz

    Kinahan
    Kinahan

    Ab

    AB2




    To:
    From:
    Subject:
    Comment:
    Link:
      
  •  
    comments powered by Disqus