Devika Subramanian, PhD
Title: Statistical learning models for real-world problems: opportunities and challenges
Abstract: In this talk, I will introduce opportunities and challenges in applying new statistical machine learning techniques to real-world problems in medicine and beyond. I will illustrate them in the context of my work on building granular predictive models for glycemic control in Type 1 diabetes, detecting lesions in early stage esophageal cancer, wind damage and street flooding risk in Houston, and inferring public sentiment toward political issues by analyzing Twitter data.
Devika Subramanian obtained a B.Tech. in Computer Science and Electrical Engineering from IIT Kharagpur and a PhD in Computer Science from Stanford. Her research areas are in artificial intelligence and machine learning, which she views as computational enablers for data-driven decision-making. She has spent the last 25 years developing novel prediction algorithms and modeling techniques for a range of decision-making problems in medicine, engineering, natural sciences, and social sciences with support from the NSF, ONR, NIH, DHS, DARPA, Microsoft, the City of Houston, and the Gulf Coast Consortia Fund. She has given several invited lectures on her work including at IJCAI and at the ONR. She has also won teaching awards at Stanford, Cornell and Rice University.
Ioannis A. Kakadiaris, PhD
Title: AI Methods In The Setting Of Precision Medicine
Professor Ioannis A. Kakadiaris, Ph.D., is a Hugh Roy and Lillie Cranz Cullen Distinguished University Professor of Computer Science at the University of Houston and directs the Computational Biomedicine Lab (CBL). He has a long and distinguished track record of research in digital health, which includes projects that: (i) Pioneered the combined use of machine learning and game theory for the prediction of adverse cardiovascular events; the ML Risk Calculator clearly outperformed the ACC/AHA Risk Calculator by recommending less drug therapy (11% versus 46% being statin eligible), yet missing fewer events (14.4% vs 23.8%); (ii) Pioneered the research field of vasa vasorum imaging for patients at risk of having a heart attack (with IP licensed to Boston Scientific), (iii) Invented and developed automated computational tools for processing large volumes of imaging data; (iv) Developed the first patient-centered software system for rendering patient-specific predictions of the post-operative breast shape for breast reconstructive surgery, (v) Invented a method and system that turns COTS mobile devices into Augmented Reality platforms able to visualize in real-time data “inside” a patient’s body (as acquired during pre-operative scans). Ioannis’ research has been supported by federal (NIH, NSF, Army Research Labs, DHS, National Institute of Justice), state (Texas Higher Education Coordinating Board), industry (SGI, American Honda, Microsoft Research, Unisys, Siemens Medical Solutions, BP America), foundations (Juvenile Diabetes Research Foundation, Schlumberger Technical Foundation), and international funding organizations (French Partner University Fund), with funds exceeding $26 million. He earned his B.Sc. in Physics at the University of Athens in Greece, his M.Sc. in Computer Science from Northeastern University and his Ph.D. at the University of Pennsylvania. In addition to twice winning the UH Computer Science Research Excellence Award, Ioannis has been recognized for his work with several distinguished honors, including the NSF Early Career Development Award, the Schlumberger Technical Foundation Award, the UH Enron Teaching Excellence Award, and the James Muller Vulnerable Plaque Young Investigator Prize. His research has been featured on Discovery Channel, National Public Radio, KPRC NBC News, KTRH ABC News, and KHOU CBS News.
Robert Grzesczuk, PhD
Title: Deep Learning for Natural Language Processing
Abstract: From Amazon Alexa to Google Translate, Natural Language Processing has been gaining prominence, driving adoption of language-based and voice-first technologies and producing rapid innovation in a dramatic shift away from conventional methods. And Healthcare has started seeing many promising applications which use these exciting new technologies. During the session, we will review a collection of representative Deep Learning methodologies and applications that leverage computer understanding and manipulation of human language in order to eliminate the tedious, time-consuming non-clinical aspects of care delivery, allowing physicians to focus on what they do best: delivering care to patients.
The session will start with an overview of word embeddings, and continue with Named Entity Recognition, Sentiment/Polarity Analysis with Convolutional Neural Networks, Question Answering and Conversational Agent systems as well as Sequence-to-Sequence Learning for Neural Machine Translation. We will also include demonstrations and proof points of clinically relevant applications, including Alexa-like virtual clinical assistants, real-time scribe for Code Blue events, digital shift handoff assistants, and quality editors for diagnostic imaging reports.
Robert is the founder and CEO of inContext.ai, a Houston based startup whose mission is to improve the quality of care and to streamline its delivery by providing clinicians with cognitive computing tools that eliminate many tedious, time-consuming non-clinical aspects of care delivery, allowing physicians to focus on what they do best: delivering care to patients.
Robert started his professional career as faculty at the University of Chicago, where he conducted research in the areas of computer graphics, scientific visualization, statistical image analysis, image segmentation & understanding, computer vision, virtual environments, computer assisted surgery (CAS), computer assisted detection (CAD), diagnosis, and treatment (surgery, radiation therapy), 3D clinical data registration and fusion, computational geometry, distributed computing, descriptive topology, and non-Euclidean geometries.
Robert is also a seasoned technical executive with more than 15 years of experience building and leading large, diverse, and globally located product development and technical organizations focused on delivering award-winning software solutions for the Healthcare Enterprise. The diverse array of companies Robert has worked for includes: Silicon Graphics (a leader in high-end computer graphics workstations), CBYON (Computer Assisted Surgical navigation tools for minimally invasive laparoscopic procedures of the spine), MagicEarth (provider of seismic data interpretation software capable of real-time roaming of Terabyte volumes in a virtual environment), Scimagix (provider of image-based content retrieval, management, and mining of visual information from pharmaceutical data for bioinformatics applications), BioImagene (provider of Digital Pathology solutions), Stentor/Philips (provider of iSite , a leading PACS solution), and Imorgon Medical (imaging solutions for integrated clinical Ultrsound).
Entrepreneurial and accomplished inventor with several valuable patents which have been acquired and commercialized by global industry leaders including Microsoft and General Electric Medical Systems. Involved in several startups with successful exits, including Stentor (acquired by Philips Medical Systems), Bioimagine (acquired by Roche), and MagicEarth (acquired by Haliburton).
Xiaoqian Jiang, PhD
Title: Challenges and Models in Protecting Biomedical Data Sharing and Analysis
Abstract: In the era of precision medicine, a lot of data are being collected from individuals to promote scientific research. On the other hand, such a massive collection of data leads to many challenges to privacy and security. Electronic Health Records, genomics, imaging, notes can be uniquely linked to individual patients to reveal their identities and put their privacy at risk, which can have profound impact to employment, insurance, etc. Techniques that are seemingly protective might not provide sufficient protection while restrictive policy can hinder the discovery in biomedicine. It is a non-trivial task to strike the right balance between privacy/security and utility to abide by law and regulation. This presentation will introduce recently revealed privacy/security problems in healthcare research and several state-of-the-art technical solutions that might throw lights to address the privacy/security risks in healthcare data sharing and analysis.
Dr. Jiang is a Christopher Sarofim family associate professor, CPRIT Scholar for Cancer Research, and Center Director for Health Security and Phenotyping in the School of Biomedical Informatics (SBMI) at The University of Texas Health Science Center at Houston (UTHealth). Before joining UTHealth in 2018, he was an associate professor with tenure in the Department of Biomedical Informatics (DBMI) at UCSD. With a Computer Science PhD from Carnegie Mellon University, he works primarily in health data privacy and predictive models in biomedicine. In the last six years, he received NIH R00, R13, R21, R01, U01, OT3 grants as PI, obtained career awards like CPRIT Rising Stars and UT Stars, and won the best and distinguished paper awards from American Medical Informatics Association (AMIA) Joint Summits on Translational Science (2012, 2013, 2016). He is a main organizer of the International Workshop on Genome Privacy and Security (GenoPri) and iDASH Genome Privacy competition (2014 - present), which was reported by Nature News and GenomeWeb.
Anthony C. Chang, MD, MBA, MPH, MS
Title:The Future of Intelligence-Based Pediatric Radiology: A Singular Moonshot
Abstract: “The important achievement of Apollo was demonstrating that humanity is not forever chained to this planet and our visions go rather further than that, and our opportunities are unlimited.“
Neil Armstrong, Apollo 11 commander
Just like the Apollo space program, the promise of artificial intelligence in medicine and health care looms large in its promise. From automating repetitive tasks with robotic process automation to interpreting medical images with convolutional neural networks as well as precision medicine with deep learning, machine intelligence with its higher accuracy and faster velocity can have impact in most if not all dimensions of pediatric radiology. There is no obvious dramatic denouement for our mission to deploy artificial intelligence in medicine and health care to signal final success, but a grand vision in pediatric radiology would be a virtual pediatric image learning system in which all pediatric medical images from all pediatric institutions are labeled and analyzed as well as rendered to continuously learn to become the most intelligent pediatric image specialist in the world.
Dr. Chang attended Johns Hopkins University for his B.A. in molecular biology prior to entering Georgetown University School of Medicine for his M.D. He then completed his pediatric residency at Children’s Hospital National Medical Center and his pediatric cardiology fellowship at the Children’s Hospital of Philadelphia. He then accepted a position as attending cardiologist in the cardiovascular intensive care unit of Boston Children’s Hospital and as assistant professor at Harvard Medical School.
He has been the medical director of several pediatric cardiac intensive care programs (including Children’s Hospital of Los Angeles, Miami Children’s Hospital, and Texas Children’s Hospital). He served as the medical director of the Heart Institute at Children’s Hospital of Orange County.
He is currently the Chief Intelligence and Innovation Officer (CIIO) and Medical Director of the Heart Failure Program at Children’s Hospital of Orange County. He has also been named a Physician of Excellence by the Orange County Medical Association and Top Cardiologist, Top Doctor for many years as well as one of the nation’s Top Innovators in Healthcare.
He has completed a Masters in Business Administration (MBA) in Health Care Administration at the University of Miami School of Business and graduated with the McCaw Award of Academic Excellence. He also completed a Masters in Public Health (MPH) in Health Care Policy at the Jonathan Fielding School of Public Health of the University of California, Los Angeles and graduated with the Dean’s Award for Academic Excellence. Finally, he graduated with his Masters of Science (MS) in Biomedical Data Science with a subarea focus in artificial intelligence from Stanford School of Medicine and has completed a certification on artificial Intelligence from MIT. He is a computer scientist-in-residence and a member of the Dean’s Scientific Council at Chapman University.
He has helped to build a successful cardiology practice as a startup company and was able to complete a deal on Wall Street. He is known for several innovations in pediatric cardiac care, including introducing the cardiac drug milrinone and co-designing (with Dr. Michael DeBakey) an axial-type ventricular assist device in children. He is a committee member of the National Institute of Health pediatric grant review committee. He is the editor of several textbooks in pediatric cardiology, including Pediatric Cardiac Intensive Care, Heart Failure in Children and Young Adults, and Pediatric Cardiology Board Review.
He is the founder of the Pediatric Cardiac Intensive Care Society (PCICS) that launched the multi-disciplinary focus on cardiac intensive care for children. He is also the founder of the Asia-Pacific Pediatric Cardiac Society (APPCS), which united pediatric cardiologists and cardiac surgeons from 24 Asian countries and launched a biennial meeting in Asia that now draws over 1,000 attendees.
He is the founder and medical director of the Medical Intelligence and Innovation Institute (MI3) that is supported by the Sharon Disney Lund Foundation. The institute is dedicated to implement data science and artificial intelligence in medicine and is the first institute of its kind in a hospital. The new institute is concomitantly dedicated to facilitate innovation in children and health care all over the world. He is the organizing chair for the biennial Pediatrics2040: Emerging Trends and Future Innovations meeting as well as the founder and director of the Medical Intelligence and Innovation Summer Internship Program, which mentors close to 100 young physicians-to-be every summer. He has organized a pediatric innovation leadership group called the International Society for Pediatric Innovation (iSPI).
He intends to build a clinician-computer scientist interface to enhance all aspects of data science and artificial intelligence in health and medicine. He currently lectures widely on big data and artificial intelligence in medicine (he has been called “Dr. A.I.” by the Chicago Tribune and has given a TEDx talk as well as on the Singularity University faculty) ([1]). He has published review papers on big data and predictive analytics as well as machine learning and artificial intelligence in medicine ([2])([3]). He is on the editorial board of the Journal of Medical Artificial Intelligence. He is currently completing a book project with a book series with Elsevier: Medical Intelligence: Principles and Applications of Artificial Intelligence in Medicine and Healthcare. He is the founder and organizing chair of several Artificial Intelligence in Medicine (AI Med) meetings in the U.S. and abroad (Europe and Asia) that will focus on artificial intelligence in healthcare and medicine (www.ai-med.io). He intends to start a new group for clinicians with a special focus on data science and artificial intelligence (tentatively MD4ai) as part of a nascent society (Medical Intelligence Society, or MIS).
He is the founder of three startup companies in artificial intelligence in medicine:
- CardioGenomic Intelligence (CGI), LLC is a multifaceted company that focuses on artificial intelligence applications such as deep learning in clinical cardiology (cardiomyopathy and heart failure as well as other cardiovascular disease) and genomic medicine.
- Artificial Intelligence in Medicine (AIMed), LLC is an events company that organizes meetings and educational programs in artificial intelligence in medicine in local as well as global venues.
- Medical Intelligence 10 (MI10), LLC is an education and consulting/advising conglomerate for executives and physician leaders as well as investors for the evaluation and implementation of AI strategies in healthcare organizations, for evaluation and recommendation of AI in healthcare vendors, and assessment and implementation of cybersecurity in healthcare organizations.
[1] https://www.youtube.com/watch?v=Y5T8nckyuCA
[2] www.congenitalcardiologytoday.com/index_files/CCT-NOV12-NA.pdf
[3] www.congenitalcardiologytoday.com/index_files/CCT-APR13-NA.pdf
Kristen Yeom, MD
Title: AI and future steps towards precise pediatric neuroradiology
Abstract: In this talk, we will examine machine learning approaches and computer vision applied to clinical pediatric neuroradiology. Topics covered will include AI applied to clinical workflow in subspecialty pediatric neuroradiology; future roles for AI in evaluating neurologic and neuro-oncologic disorders of childhood that combine precision medicine and imaging; and AI methods for safer and more efficient neurosurgical navigation. Brief discussion of current literature and various clinical AI models will be presented, as well as how a radiologist can start building AI clinical models or utilize currently available models for education, research, or for augmentation of clinical skills.
Kristen Yeom is an Associate Professor of Radiology and faculty at the Center for Artificial Intelligence in Medicine and Imaging at Stanford University. She specialized in neuroradiology and serves as the interim director of Pediatric Neuroradiology and Associate Director of MRI at Lucile Packard Children’s Hospital at Stanford. She obtained medical degree from University of Michigan, Diagnostic Radiology residency at UCLA School of Medicine, and Neuroradiology fellowship at Stanford University. Dr. Yeom’s research has focused on clinical and translational studies of advanced MRI methods, such as diffusion, perfusion, and quantitative susceptibility MRI, as well as novel image processing tools for improved understanding of normal neural development and diagnosis and management of neurological and neuro-oncologic diseases. Her recent works include radiomic and machine-learning strategies for pediatric brain tumor classification, as well as computer vision tasks for clinical neuroimaging diagnostics, such as deep neural networks for assessing normal and abnormal brain development and aging, and creation of deep vision classifier models for brain and neurovascular pathologies.
Luca Giancardo, PhD
Abstract: Medical image computing applications using machine learning approaches present challenges that are often unique to the medical world. In this talk, I will present a range of applications in radiology using off the shelf deep learning models and approaches specifically tailor for medical images. Specifically, we will cover applications for safety of implantable devices using x-rays, longitudinal brain analysis for neurodegerative diseases and computer aided decision support for stroke.
Luca Giancardo is an Assistant Professor at the Center for Precision Health, School of Biomedical Informatics (SBMI), UTHealth with co-appointments at and at the Diagnostic and Interventional Imaging, McGovern Medical School, UTHealth and the Institute for Stroke and Cerebrovascular Diseases, UTHealth. He was a research fellow at the Massachusetts Institute of Technology, Postdoc at the Italian Institute of Technology (Italy). He received his PhD at the Oak Ridge National Laboratory and University of Burgundy (France).
He is a computer scientist with extensive experience in image analysis and machine learning. He has worked on developing new machine learning-based methodologies to discover computational biomarkers from patterns in biomedical data such as optical images, magnetic resonance imaging, X-rays, computer tomography, laboratory animal videos or tracking devices. His work has been applied to a number of biomedical applications, such as diabetic retinopathy screening or neurodegenerative disease tracking and successfully translated to industry with two startups based on his methods. One, Hubble Telemedical, was acquired by Welch Allyn in 2015, another, nQ-Medical has raised multimillion dollars in investment in 2018. His work has been featured by news outlet such as MIT Technology Review, Smithsonian magazine and others. I have received multiple awards, including the prestigious 100k Singapore Challenge awarded by a judging panel composed by Nobel Prize and Millennium Technology Prize winners.
Matthias Wagner, MD
Title: From nice-to-have to must-have. Selected AI applications in pediatric neuroradiology.
Abstract: Outcome prediction of premature brain injury and detection of molecular markers of low grade gliomas are two examples of promising yet highly specific AI assisted applications in pediatric neuroradiology. Deep learning algorithms that allow to predict cognitive, language, and motor outcome at 4.5 years of age and that support discovering genomic markers are features that are a “nice-to-have” but not a “must-have” product. They might even reduce radiologic workflow by inducing additional clicks and visual analysis efforts and they might even divert attention from subtle findings. What needs to be proven for must-have products is that AI assisted prediction or clinical decision support tools lead to a better patient outcome and to a reduction of healthcare costs. One example that may improve patient care and help reducing costs is an objective, accurate, and robust method to predict abnormal imaging findings. With the increasing number of patients in emergency departments, we expect to see further increase of neuroimaging studies request by emergency physicians. However, technical resources and personnel including MRI technicians and radiologists are limited. Therefore, it becomes increasingly challenging for diagnostic imaging departments to accommodate these requests. Consequently, there is a need to develop a model that can be applied by emergency physicians to objectively classify patients who have a greater risk of abnormal findings on their brain MRIs – in other words: a model that groups patients into an “imaging needed” and “imaging not needed” category.
Matthias Wagner is Chief Fellow in the Department of Diagnostic Imaging at The Hospital for Sick Children, Toronto. During his radiology residency at the Zurich University Hospital, Switzerland, he worked as a research fellow at The Johns Hopkins Hospital, Division of Pediatric Neuroradiology. After the radiology board exams in 2018, he completed a fellowship in pediatric neuroradiology at SickKids. Currently, Matt is leading the radiology AI research at SickKids and is working on a Master’s degree in Computer Science.
Zbigniew Starosolski, PhD
Title: Data Security in the context of AI's for imaging
Dr. Zbigniew Starosolski earned his Ph.D. in the field of Control Theory at the Faculty of Automatic Control of the Silesian University of Technology (Gliwice, Poland) in 2004. He worked as an Assistant Professor at his alma mater for 11 years, teaching a series of lectures on Theory of Automatic Control, Dynamical Systems and Biostatistics. During that period, he also completed a Postdoctoral Fellowship at School of Biomedical Informatics at The University of Texas Health Science Center in Houston in 2011. Dr. Starosolski joined the Department of Radiology, Texas Children’s Hospital as a Senior Research Scientist in 2011. Since 2013 he has been an Assistant Professor at Department of Radiology, Baylor College of Medicine.
His research interests in addition to the applications of machine learning in medicine, include applying control theory to molecular dynamics of protein interaction and rapid prototyping for clinical applications. He developed several algorithms for image recognition and classification of pre-clinical magnetic resonance images and computed tomography images. He developed the image analysis method using radiomics features for tumor aggressiveness assessment in animal models of pediatric tumors. His recent work is focused on machine learning approaches using Deep Convolutional Neural Networks applied to pediatric clinical data, including imaging and EHR. These indicate that carefully trained and characterized AI models could be a useful tool for an early detection of several conditions in pediatric patients.
John Hamm
John Hamm is a strategist fueled by the idea that technology can change the way people feel, connect, and work. His career started after high school implementing computers and coding at a global car company. Growing up in Dayton, Ohio, John and family found their way to Houston, TX in the late-1990s.
John’s accomplishments include executing a hybrid data center transformation at Texas Children's Hospital, establishing a high-speed artificial intelligence in-memory platform on SAP HANA at a Fortune 100 oil and gas company, building an IT organization from the ground up at a 6 billion dollar chemical distributor, and leading global SAP mergers at a fortune 100 global chemical manufacturer. John establishes bold strategies and produces results, embracing and empowering competent people along the journey.
John completed his Master of Business Administration degree while raising a family of four boys. He welcomed his little girl in 2016 at Texas Children’s! The experience reaffirmed his passion for improving women and children's lives by enabling caregivers with technology and data at Texas Children's; a multi-billion dollar integrated health system. Obsessed with people and technology, he pursues excellence in connecting with and developing his teams, staying affluent with cutting edge technology, and remaining in the line of sight with what happens on the front lines.
Bill O’Donnell
Title: Medical Data Security
Abstract: Artificial Intelligence algorithms require a massive amount of data to be effective, must of which comes from consumers. In the health industry, a delicate balance exists between data privacy from patient data and providing machine learning diagnosis predictions. A significant portion of data privacy protection revolves around regulatory compliance. HIPAA and potentially GDPR are regulatory standards that must be followed in the health industry. There are many aspects to consider when using AI applications in viewing or processing sensitive data. This session will review best practices when dealing with security and regulatory compliance protecting the privacy of data./p>
Bill O’Donnell is Nutanix Principal Security Product Management. With over 25 years’ experience in information technology, Bill has been involved in security design, architecture, and product management of a wide range of challenging security solutions in software and hardware products. Formally worked as a senior security leader in IBM spearheaded secure, flexible, and cutting-edge solutions, where he was the chief security architect for IBM WebSphere Brand as well as authored several security papers and blogs including the company’s Security Intelligence blog.