Medical imaging started probably 123 years ago when German physicist Wilhelm Röntgen showed his wife an X-ray of her hand. At that time, Anna Röntgen was so overwhelmed with what she saw that she thought she had just witnessed death.
Today, “we are at a point where the idea that we can do AI for medical imaging should not be at all weird or controversial”, said Jeremy Howard, founding research of fast.ai at last year’s AI med North America. Indeed, AI involving imaging and diagnostics peaked in 2015 and still holds a steady growth.
However, data storage is challenging the development. As computed tomography (CT) scan, Magnetic resonance imaging (MRI), 3D and 4D imageries became competent to capture the thinnest slice of our body, hospitals are producing 50 petabytes of data every year. Like data accumulated from electronic medical records, up to 97% of these images go without used or further analyzed after diagnosis and treatment.
Cloud computing has offered an astute solution, encouraging physicians to gather, manage or even transfer information remotely but coming at a price. Without high speed internet, systematic data protection and clarify who – patients, doctors or hospitals – owns the data, the use of Cloud computing to establish collaboration, remains ideal.
On the other hand, clinicians and developers are testing the limits of deep learning; how it’s able to permit early discovery of a medical condition. With YOLO, a real time detection and classification tool, driven by Convolutional Neural Networks (CNNs) and Graphic Processing Unit (GPU), researchers were able to perform segmentation by having the program to review patients’ MRI and calculates the volume left ventricle within each diastole and systole end frame and thereby determining an individual’s heart health. A task which was performed manually by a radiologist manually.
In this session, Google AI and one of its three partnering universities will be sharing some of their breakthroughs in medical imaging; how computer vision has learn to achieve the kind of identification accuracy no difference from an experienced physician.
For more information on this topic, register for AIMed North America here.
Session Focus: Medical Imaging: Cardiology/Pathology/Radiology
When: Friday, December 14th 2018 (13.00-14.00)
An overview of recent medical imaging breakthroughs; how have AI, specifically deep learning and computer vision aided its overall developments.
Attendees will gain the following knowledge:
Discover the latest innovations of medical imaging and the kind of challenges that these novelties had posed to the sector.
Be informed of the next medical imaging trends, what are the kind of collaboration and advances which clinicians and developers are seeking after.
Learn from the sector’s forerunners; how will the medical imaging becomes more effective and efficient, with the help of AI.
Benefit from an afternoon of intellectual exchange and an in-depth understanding of a subspecialty field.
Lawrence Tanenbaum, Vice President and Director of Advanced Imaging, Radnet, USA
Dennis Wall, Associate Professor of Pediatrics, Stanford University, USA
Peter Chang, Director, Center for AI in Diagnostic Medicine, UC Irvine Health and CEO, Avicenna.ai, USA
Afshin Aminian, Medical Director, Orthopaedic Institute, CHOC, USA
Register for AIMed North America here for this session.