What does the future of artificial intelligence in medicine look like? Will AI become the clinician’s best friend or replacement? Can AI change the way we conduct clinical trials, drug development and hospital operations, or is the path forward unclear? At this year’s World Medical Innovation Forum, held April 23 to 25, BWH thought leaders and rising stars joined took the stage to address these questions and more. This year’s sold out forum had 1,700 registrants, including 140 executive presenters and key leaders. If you were unable to attend — or if you want to revisit highlights from this year’s forum — check out CRN’s coverage in the images and extended captions below.

Disruptive Dozen

AI rock stars from BWH and MGH revealed the top 12 AI technologies that will reinvent health care over the next decade.

1: AI at the Bedside: Brandon Westover, MD, PhD, Director, MGH Clinical Data Animation Center, MGH. Westover discusses the potential for artificial intelligence to continuously mine bedside data, such as blood pressure readings, EEGs or ECGs, in an automated fashion to forecast adverse events, helping doctors monitor patients who are at a particularly high risk.
1: AI at the Bedside: Brandon Westover, MD, PhD, Director, MGH Clinical Data Animation Center, MGH. Westover discusses the potential for artificial intelligence to continuously mine bedside data, such as blood pressure readings, EEGs or ECGs, in an automated fashion to forecast adverse events, helping doctors monitor patients who are at a particularly high risk.
2: A Picture is Worth a Thousand Words: Hadi Shafiee, PhD, Director, Laboratory of Micro/Nanomedicine and Digital Health, BWH. In medicine, an image is more than just a picture; it is millions, even billions of data points that can be systematically scrutinized and mined for connections to health a disease. Shafiee discusses how artificial intelligence is transforming how these digital images are analyzed and interpreted in various spheres of health care. The majority of the population is equipped with pocket-sized, powerful devices with sensors built inside – this is a great opportunity Every day in our digital world, we generate 2.5 TB of data Outside of health care, there is something big happening in the world of smartphones and we can leverage that technology to address important problems in disease management at the point of care
2: A Picture is Worth a Thousand Words: Hadi Shafiee, PhD, Director, Laboratory of Micro/Nanomedicine and Digital Health, BWH. In medicine, an image is more than just a picture; it is millions, even billions of data points that can be systematically scrutinized and mined for connections to health a disease. Shafiee discusses how artificial intelligence is transforming how these digital images are analyzed and interpreted in various spheres of health care. The majority of the population is equipped with pocket-sized, powerful devices with sensors built inside – this is a great opportunity Every day in our digital world, we generate 2.5 TB of data Outside of health care, there is something big happening in the world of smartphones and we can leverage that technology to address important problems in disease management at the point of care
3: Can Personal Devices Improve Your Health? Omar Arnaout, MD, Co-Director Computation Neuroscience Outcomes Center, Attending Neurosurgeon, Department of Neurosurgery, BWH. Digital devices permeate our lives. Arnaout talks about how our personal devices, such as smartphones, Fitbits and Apple Watches, might become another tool in the clinician’s toolbox to help predict disease and improve health. Challenge is getting people to agree to collect their personal data constantly collected Patients are likely to sign up to help patients who may benefit down the line
3: Can Personal Devices Improve Your Health? Omar Arnaout, MD, Co-Director Computation Neuroscience Outcomes Center, Attending Neurosurgeon, Department of Neurosurgery, BWH. Digital devices permeate our lives. Arnaout talks about how our personal devices, such as smartphones, Fitbits and Apple Watches, might become another tool in the clinician’s toolbox to help predict disease and improve health. Challenge is getting people to agree to collect their personal data constantly collected Patients are likely to sign up to help patients who may benefit down the line
4: Risky Business: Using EHRs to Predict Disease Risk: Ziad Obermeyer, MD, Assistant Professor, Emergency Medicine, BWH. Making health care more precise is a holy grail of modern medicine. Obermeyer discusses how AI, specifically machine and deep learning, can harness “big data” made available by electronic health records – enabling a new wave of biomedical and clinical research, generating knowledge to help realize the goals of so-called precision medicine. Data is the new oil – it’s cleaner, renewable but it doesn’t come right out of the ground. It requires work to get it. The term “data mining” – I work in a data mine. And it’s hard work. You may be predicting billing for a stroke rather than risk of disease The biggest challenge may be understanding exactly what are we predicting with our models
4: Risky Business: Using EHRs to Predict Disease Risk: Ziad Obermeyer, MD, Assistant Professor, Emergency Medicine, BWH. Making health care more precise is a holy grail of modern medicine. Obermeyer discusses how AI, specifically machine and deep learning, can harness “big data” made available by electronic health records – enabling a new wave of biomedical and clinical research, generating knowledge to help realize the goals of so-called precision medicine. Data is the new oil – it’s cleaner, renewable but it doesn’t come right out of the ground. It requires work to get it. The term “data mining” – I work in a data mine. And it’s hard work. You may be predicting billing for a stroke rather than risk of disease The biggest challenge may be understanding exactly what are we predicting with our models
5: Reading the Tea Leaves of Cancer Immunotherapy: Long Le, MD, PhD, Director, Computational Pathology and Director, Technology Development, Center for Integrated Diagnostics, MGH. Over the last few years, cancer immunotherapies have yielded impressive outcomes in combating some forms of the disease, but the majority of patients do not respond. Since these treatments come with high risk of severe side effects and high costs, physicians hope to target these new therapies to patients who are most likely to respond. Le discusses AI-based approaches to identify patients for whom immunotherapies will be most effective.
5: Reading the Tea Leaves of Cancer Immunotherapy: Long Le, MD, PhD, Director, Computational Pathology and Director, Technology Development, Center for Integrated Diagnostics, MGH. Over the last few years, cancer immunotherapies have yielded impressive outcomes in combating some forms of the disease, but the majority of patients do not respond. Since these treatments come with high risk of severe side effects and high costs, physicians hope to target these new therapies to patients who are most likely to respond. Le discusses AI-based approaches to identify patients for whom immunotherapies will be most effective.
6: Bringing “Smart” Machines to Medicine: Mark Michalski, MD, Executive Director, MGH & BWH CCDS. Michalski reports on the efforts to engineer a new generation of “smart” medical equipment that melds scanning with interpretation, monitoring with treatment, and merges data from disparate devices into a common, readily interpretable stream. Opportunity to make impacts where human can’t AI can integrate disparate bits of information from across the health care system to alert a clinician to intervene early on Raw data from scanners may include biomarkers we can’t see The potential of smart devices is to be able to do what humans cannot and the ability to take humans out of the equation when a person is not available
6: Bringing “Smart” Machines to Medicine: Mark Michalski, MD, Executive Director, MGH & BWH CCDS. Michalski reports on the efforts to engineer a new generation of “smart” medical equipment that melds scanning with interpretation, monitoring with treatment, and merges data from disparate devices into a common, readily interpretable stream. Opportunity to make impacts where human can’t AI can integrate disparate bits of information from across the health care system to alert a clinician to intervene early on Raw data from scanners may include biomarkers we can’t see The potential of smart devices is to be able to do what humans cannot and the ability to take humans out of the equation when a person is not available
7: Harnessing the Power of Digital Pathology :Jeffrey Golden, MD, Chair, Department of Pathology, BWH. Golden gives us a glimpse of a new generation of AI-based, data-driven methods in pathology, that can help clinicians tame complexity and make more precise diagnoses and outcome predictions for their patients. Value: value is around what pathology delivers 70 percent of all decisions in health care are based on pathology results The more accurate we do it, the sooner we get to the right diagnosis, the better we’re going to be Efficiency: the pathologist is being asked to do more work – AI systems can direct us at the right thing to look at so that we can assess what’s important and what’s not when making a decision for each patient
7: Harnessing the Power of Digital Pathology :Jeffrey Golden, MD, Chair, Department of Pathology, BWH. Golden gives us a glimpse of a new generation of AI-based, data-driven methods in pathology, that can help clinicians tame complexity and make more precise diagnoses and outcome predictions for their patients. Value: value is around what pathology delivers 70 percent of all decisions in health care are based on pathology results The more accurate we do it, the sooner we get to the right diagnosis, the better we’re going to be Efficiency: the pathologist is being asked to do more work – AI systems can direct us at the right thing to look at so that we can assess what’s important and what’s not when making a decision for each patient
8: Minimizing the Threats of Antimicrobial Resistance and Infections Associated with Antibiotic Use: Erica Shenoy, MD, PhD, Associate Chief, Infection Control Unit, MGH. Shenoy reports on the decline in usefulness of antibiotics and troubling consequences, such as the rise in infections with multi-drug resistant organism “super bugs” and the danger of infections linked to antibiotic use. Researchers are now leveraging machine learning and other artificial intelligence approaches to predict which patients may be most vulnerable to such infections and develop target interventions to minimize the dangers posed by such formidable bacteria.
8: Minimizing the Threats of Antimicrobial Resistance and Infections Associated with Antibiotic Use: Erica Shenoy, MD, PhD, Associate Chief, Infection Control Unit, MGH. Shenoy reports on the decline in usefulness of antibiotics and troubling consequences, such as the rise in infections with multi-drug resistant organism “super bugs” and the danger of infections linked to antibiotic use. Researchers are now leveraging machine learning and other artificial intelligence approaches to predict which patients may be most vulnerable to such infections and develop target interventions to minimize the dangers posed by such formidable bacteria.
9: Getting Back to Face Time: AI Tools that Help Reduce Physicians’ Computer Use: Adam Landman, MD, VP and CIO, Brigham Health. For each hour of face-to-face interaction with patients, physicians log almost two hours on administrative computer tasks. Landman talks about how artificial intelligence and machine learning promise to help lighten physicians’ computer workload by automating tasks, leaving more time for face-to-face interactions with patients. There’s an opportunity now to rethink the EHR to make it more patient focused and improve the physician experience Clinical documentation – you can type notes but let’s consider being more bold and video record and index for future retrieval Virtual assistants to be brought to the bedside In-basket remains time consuming – AI has the opportunity to take over the mundane tasks and bring to the top of the in-basket the topics to focus on
9: Getting Back to Face Time: AI Tools that Help Reduce Physicians’ Computer Use: Adam Landman, MD, VP and CIO, Brigham Health. For each hour of face-to-face interaction with patients, physicians log almost two hours on administrative computer tasks. Landman talks about how artificial intelligence and machine learning promise to help lighten physicians’ computer workload by automating tasks, leaving more time for face-to-face interactions with patients. There’s an opportunity now to rethink the EHR to make it more patient focused and improve the physician experience Clinical documentation – you can type notes but let’s consider being more bold and video record and index for future retrieval Virtual assistants to be brought to the bedside In-basket remains time consuming – AI has the opportunity to take over the mundane tasks and bring to the top of the in-basket the topics to focus on
10: Disseminating Medical Expertise to Areas that Need it Most: Jayashree Kalpathy-Cramer, PhD, Assistant in Neuroscience, MGH. Kalpathy-Cramer talks about how artificial intelligence may help to fill the gap in clinical expertise in regions where doctors are in short supply. For example, there is a global shortage of radiologists and an increasing patient population seeking CT and MRI scans, and harnessing deep-learning methods may allow for automated tools to process such scans, filling the gap.
10: Disseminating Medical Expertise to Areas that Need it Most: Jayashree Kalpathy-Cramer, PhD, Assistant in Neuroscience, MGH. Kalpathy-Cramer talks about how artificial intelligence may help to fill the gap in clinical expertise in regions where doctors are in short supply. For example, there is a global shortage of radiologists and an increasing patient population seeking CT and MRI scans, and harnessing deep-learning methods may allow for automated tools to process such scans, filling the gap.
11: Next-Gen Radiology: Alexandra Golby, MD, Neurosurgeon, Director of Image guided Neurosurgery, BWH. Golby discusses the emerging field of radiomics, which harnesses image-based algorithms to visually and non-invasively characterize tumor phenotypes and genetic makeup, which could enable more rapid diagnoses and improved treatments. Could we predict the aggressiveness of prostate cancer from an image rather than a biopsy? Researchers across the world are leveraging AI in pursuit of this goal The challenge: to bridge different techniques at different scales and with different teams We want to bring together diagnosticians, radiologists, surgeons and pathologists, says Golby
11: Next-Gen Radiology: Alexandra Golby, MD, Neurosurgeon, Director of Image guided Neurosurgery, BWH. Golby discusses the emerging field of radiomics, which harnesses image-based algorithms to visually and non-invasively characterize tumor phenotypes and genetic makeup, which could enable more rapid diagnoses and improved treatments. Could we predict the aggressiveness of prostate cancer from an image rather than a biopsy? Researchers across the world are leveraging AI in pursuit of this goal The challenge: to bridge different techniques at different scales and with different teams We want to bring together diagnosticians, radiologists, surgeons and pathologists, says Golby
12: Melding Mind and Machine: Leigh Hochberg, MD, PhD, Director, Center for Neurotechnology and Neurorecovery, Neurology, MGH. While still experimental, brain-computer interfaces are ushering in a new era of AI-based neurotechnologies that promise to restore the power of mobility and communication to thousands of people with debilitating neurological conditions. Hochberg discusses the successes of brain-computer interfaces thus far, and how artificial intelligence is propelling the development of such devices.
12: Melding Mind and Machine: Leigh Hochberg, MD, PhD, Director, Center for Neurotechnology and Neurorecovery, Neurology, MGH. While still experimental, brain-computer interfaces are ushering in a new era of AI-based neurotechnologies that promise to restore the power of mobility and communication to thousands of people with debilitating neurological conditions. Hochberg discusses the successes of brain-computer interfaces thus far, and how artificial intelligence is propelling the development of such devices.

 

First Look: The Next Wave of AI Breakthroughs in Health Care

Nine early-career investigators from the Brigham kicked off the 2018 World Medical Innovation Forum with rapid-fire presentations on their bold ideas for leveraging artificial intelligence to advance medicine and patient care. Their ten-minute presentations highlight discoveries and insights that may shape the future of health care.

Nathalie Agar, PhD presented “Mass Spectrometry Imaging for Surgical Pathology and Oncology.” Agar, who directs the Surgical Molecular Imaging Laboratory in the Department of Neurosurgery, is developing new techniques that uses mass spectrometry imaging to help surgeons more precisely define the locations of tumors in near real-time.
Ziad Obermeyer, MD, MPhil – A Machine Learning Algorithm to Reduce Costs and Increase Quality
Hadi Shafiee, PhD – AI-Empowered Optical System for In-Vitro Fertilization
Omar Arnaout, MD – Using Residual Convolutional Neural Networks to Improve Treatment of Brain Tumors
Currently, the only way to determine if a glioblastoma patient has an IDH-mutation, a mutation resulting in significantly longer survival times, is through invasive surgical procedures. Omar Arnaout MD, discusses how his group is using residual convolutional neural networks to predict whether patients have this mutation and others without such invasive procedures.
Michael Rosenthal, MD, PhD – Extracting Muscle and Fat Metrics from Clinical CT Images Using Neural Networks
One of these models was created manually and one was automated. “Can you tell the difference?”Rosenthal asks. Rosenthal, a radiologist in the Department of Radiology, focuses his clinical work on the imaging of gastrointestinal cancers. His research uses machine learning techniques to understand pancreas cancer metabolism via analysis of body fat and muscle on CT scans.
Agar is leveraging the cutting-edge resources of the AMIGO suite, pictured here, to characterize tissue in real time.
Obermeyer, an assistant professor of emergency medicine and health care policy, is using the Brigham’s large, data-rich electronic health records and applying machine learning to make predictions on health outcomes, with the goal of improving clinical decision making and population health management.
Shafiee, director of the laboratory of micro/nanomedicine and digital health, is working on integrating medicine with micro- and nanotechnology to develop treatment monitoring for infertility.
Raul San Jose Estepar, PhD – Identification of Healthcare Risks in COPD Patients Using CT and X-ray Images Raul Estepar Estepar, the co-director of the Applied Chest Imaging Laboratory at BWH, is developing computational techniques to find biomarkers for Chronic Obstructive Pulmonary Disease (COPD), Pulmonary Vascular Disease (PVD) and Interstitial Lung Disease (ILD) that can be identified from radiology images.
Fedorov of the BWH Department of Radiology, is leveraging medical image computing technology in clinical research applications, with the focus on quantitative imaging, imaging informatics and image-guided interventional procedures.
William Lane, MD, PhD – bloodTyper: Blood Group Typing using Next Generation Sequencing
Lane describes the limitations of current blood typing strategies and the transfusion complications that can result from mismatches William Lane As a pathologist and director of Clinical Laboratory Informatics and assistant director of the Tissue Typing Laboratory at BWH, Lane is acutely aware of the importance of the plethora of blood antigens that differ from person to person. Blood types go well beyond ABO – hundreds of antigens exist and mismatches between donors and recipients can have repercussions for people who receive multiple blood transfusions and are also important in prenatal care.
Georg Gerber, MD, PhD – Machine Learning Tools for Microbiome Diagnostics and Therapeutics Development
Gerber, a computer scientist, microbiologist, physician and pathologist, is interested in using big data and computational models to understand the role of the microbiome in human disease.
Raul San Jose Estepar, PhD – Identification of Healthcare Risks in COPD Patients Using CT and X-ray Images
Andrey Fedorov, PhD – Data Standardization for Advancing the Role of AI in Radiology

 

Panel Discussions & Fireside Chats

BWHers led panel discussions on how artificial intelligence is transforming health care.

Brigham Health President Betsy Nabel, MD, led a fireside chat with Atul Gawande, MD, executive director of Ariadne Labs.
AI and the Cost of Trials: The Impact of Real World and Real Time Evidence Moderator: Krishna Yeshwant, MD Panelist: Stephen Wiviott, MD. Krishna Yeshwant, MD, a BWH physician and partner at Google Ventures, led a roundtable on realistic timelines for AI technology to be implemented in health care. Panelists identified bottlenecks in applying AI at a scale that will have a measurable effect on trial cost and time to conduct the trial. Clinical trials have always had mathematical and statistical tools at their heart. The rise of AI promises to increase the efficiency of clinical trials and decrease costs. Yeshwant asked the panel: What parts of the clinical trial process can be improved by AI? How can we implement AI in clinical trials? What are the risks of implementing AI in clinical trials? The panel included Stephen Wiviott, MD, an attending cardiologist at BWH who is currently directing a clinical trial on the cardiovascular safety and efficacy of metabolic therapies and is also a senior investigator for TIMI. Also serving as the executive director of the Clinical Trials Office at PHS, Wiviott is very familiar with the clinical trials process and discussed places where AI could be implemented to decrease costs and increase efficiency in the clinical trial process.
Can AI-Based Drug Development Feed A Hungry Pipeline? Moderator: Anthony Philippakis, MD, PhD Philippakis, a cardiologist at BWH, led a discussion with leaders in industry and academia about AI-based approaches to target discovery, drug design and drug repurposing. Philippakis asks, Have we seen a drug go into the clinic where machine learning was the difference between success and failure? Why has drug development lagged? Is it because we don’t have the right data, is it because we don’t have the right people, is it because there’s something different about this problem?
Machine Learning in Image Analysis: A Diagnostician’s Best Friend…or Replacement? Moderator: Giles Boland, MD. Boland, chair of the BWH Department of Radiology, leads a discussion about diagnostic imaging and the evolving role of the radiologist. Will AI tools ever become advanced enough to make decisions autonomously within the clinical workflow?
Reflecting on the Impact of AI at the Bed and the Bench: Chairs Roundtable Calum MacRae, MD, PhD. MacRae, vice chair for Scientific Innovation in the Department of Medicine, joined a panel of senior leaders, including chairs and former chairs, to discuss where AI is today and its impact on the clinic as well as bench research.