AAPM Foster July 2009

1,145 views

Published on

I gave this talk in the "Presidential Symposium" at the annual meeting of the American Association of Physicists in Medicine, in Annaheim, California. The President of AAPM, Dr. Maryellen Giger, wanted some people to give some visionary talks. She invited (I kid you not) Foster, Gates, and Obama. Fortunately Bill and Barack had other commitments, so I did not need to share the time with them.

1 Comment
1 Like
Statistics
Notes
  • After you got all the information on Fioricet, another point on your agenda should be the price for it. http://www.fioricetsupply.com resolves this problem. Now you can make the decision to buy.
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here
No Downloads
Views
Total views
1,145
On SlideShare
0
From Embeds
0
Number of Embeds
4
Actions
Shares
0
Downloads
36
Comments
1
Likes
1
Embeds 0
No embeds

No notes for slide
  • Medicine is approaching a profound transition as the methods of molecular medicine start to transform the nature of health care.What is the significance of such methods? For the researcher, it is a paradigm shift, as the number of things that can be measured increases dramatically.
  • Researchers express a vision for a scientific revolution in health care, from the qualitative to the quantitative-- A revolution based on information and thus computing
  • However, even as we talk about transformation and revolution, we must recognize that computing is poorly used in health care today.These are the words of a recent National Research Council report.Thus, I will seek in my remarks today to shed light on three questions: how information technology is evolving, how this evolution may impact medicine, and how changes in medicine and health care will stress information techology.
  • The story of computers is one of exponentials
  • The story of computers is one of exponentials
  • The story of computers is one of exponentials
  • But things are not quite as bad as that
  • What does this mean for medicine?We will certainly continue to see increasingly sophisticated computer applications aiding the physician in their tasks of observing, diagnosing, and treating – what used to be solely the domain the human senses, the brain, and the hands.More accurate, higher resolution, and more automated data acquisition systems.Computer-aided diagnosis and treatment planning systems that use large-scale data analysis and computer simulations.Automated radiation treatment and surgery systems. However, I want to focus here on some larger systems issues relating to quantitative medicine.
  • Using gene expression microarrays, we find that these two diseases have quite different phenotypes—that quite different genes are expressed in the two conditions.Here, columns are patients; rows are genes.Not sure what is the significance of the Stage 1/Stage 2.”The beauty of gene expression profiling data is that it is quantitative and highly reproducible. Because of this, these data can be used to generate multivariate statistical models of the clinical behavior of cancer that have great predictive power.” -- http://lymphochip.nih.gov/Staudt_Adv_Immunol_2005.pdf
  • And of course, we must not forget image-based biomarkers, as used in computer aided diagnosis of breast cancer, or as shown here, in an attempt to identify biomarkers for traumatic brain injury.ROIs used in a study at UIC(A) forceps minor (green), cortico-spinal tract (purple), inferior frontal-occipital fasciculus (red), external capsule (yellow), sagittal stratum (blue) (B) anterior corona radiata (green), superior longitudinal fasciculus (red), posterior corona radiata (blue); (C) cingulum (red), corpus callosum body (blue), splenium (yellow), and genu (green), and forceps major (purple).
  • Then, by tracking the personalized treatment plan, we collect more patient data.Success demands that we integrate, to a far greater degree than previously possible, clinical practice, basic research, and clinical trials. A profound challenge for health care system and for information technology.
  • Collecting and managing the enormous quantities of data that are now feasible, and required for EBM, is a huge challenge.However, merely putting in place the systems required to collect large quantities of data is not enough.We then need to make sense of that data. A challenge both for the physician and the researcher.
  • These problems arise at multiple scales. E.g. …
  • What these (and other examples that we will not have time to review) have in common …
  • We cite [Rouse, Health Care as a CAS: Implications for Design… , NAE 2008] for the righthand side aprt.Must supportDynamic composition for a specific purposeEvolving community, function, environmentMessy data, failure, incomplete knowledgeNice, but insufficientData standardsPlatform standardsFederal policies
  • Another perspective on the problem. A few words of explanation. If we are deploying a hospital IT system, we are (hopefully) in the bottom left hand corner.“You can’t achieve success via central planning.” Quoted in Crossing the Quality Chasm, p. 312In our scenarios, we don’t have that ability to control.
  • What is the alternative? We can put in place mechanisms that facilitate groups with some common goal to form and function.Over time, things change, these groups evolve.If we are successful, they can expand, perhaps merge.Challenges: make this easy. Leverage scale effects.
  • These are issues that the grid community has been working on for many years. We call these groupings Virtual Organizations.In healthcare today, there are of course many such “VOs.”But they are hard to form, fragmented, …
  • Principles and mechanisms that has been under development for some years.First CS, then physical sciences, then biology, most recently biomedicine –
  • What are these grid mechanisms and concepts, then? Hard to say something sensible in a few minutes.But basically it is about separating out concerns in a way that reduces barriers to entry and permits flexible use.
  • API vs. protocol? “Illities”?
  • [Create an image here.]For example DICOM and HL7 combine messaging and data model in the same interoperability standard. People are contextualizing this problem at the data interoperability level.  Systems interoperability often neglected.  An area of differentiation, bringing in best practice in industry and science into health care space. Open source platform.  Experience with systems interoperability standards: IETF, OASIS, W3C, 
  • Scaling via automating data adaptersRepresentations of those things and semantics of those representations.Talk about how services are published, data modeling, etc.Publish data basesPublish servicesName published objects
  • Loose coupling and encapsulationInteroperability through integration based on data mediation Evolutionary in nature Set of scalable systems and methods Explicit in architecture – data integration layerDemonstrated in GSI, GridFTP, MDS, ECOG
  • Most images are never seen—and are not available—outside their originating institution
  • Recap …
  • 6 representative challenges
  • Data sharing is the most important
  • AAPM Foster July 2009

    1. 1. The present and future role of computers in medicine<br />Ian Foster<br />Computation Institute<br />Argonne National Lab & University of Chicago<br />
    2. 2. Credits<br />Thanks for support from<br />Chan Soon-Shiong Foundation<br />Department of Energy<br />National Institutes of Health<br />National Science Foundation<br />And for many helpful conversations, Carl Kesselman, Jonathan Silverstein, Steve Tuecke, Stephan Erberich, Steve Graham, Ravi Madduri, and Patrick Soon-Shiong<br />
    3. 3. Biology is shifting from being an observational science to a quantitative molecular science<br /> Old biology: measure one/two things in two/three conditions<br />High cost per measurement<br />Analysis straightforward as little data<br />Enormously difficult to work out pathways due to inadequate data<br /> New biology: measure 10,000 things under many conditions<br />Low cost per measurement<br />Analysis no longer straightforward<br />Payoff can be bigger: potential to understand a complex system<br />Ajay Jain, UCSF<br />
    4. 4. Change health care <br />from <br />an empirical, qualitative systemof silos of information<br />to a model of<br />predictive, quantitative, shared,evidence-based outcomes <br />
    5. 5. The health care information technology chasm<br /> Health care IT [is] rarely used to provide clinicians with evidence-based decision support and feedback; to support data-driven process improvement; or to link clinical care and research.<br />Computational Technology for Effective Health Care, NRC, 2009<br />
    6. 6.
    7. 7.
    8. 8.
    9. 9. Digital power =<br />computing x communicationxstorage x content<br />Moore’s law<br />doubles<br /> every 18<br /> months<br /> disk law<br />doubles<br />x every 12<br /> months<br /> fiber law<br />doubles<br />xevery9<br /> months<br />community law<br />n<br />x 2<br /> where n is # people<br />John SeelyBrown<br />
    10. 10. (Intel)<br />
    11. 11. Microprocessor trends<br />AMD<br />Dual core (April 2005)<br />Quad core (October 2007)<br />Intel<br />Dual core (July 2005)<br />Quad core (December 2006)<br />Sun<br />Niagara: 8 cores * 4 threads/core (November 2005)<br />Niagara2: 8 cores * 8 threads/core (August 2007)<br />IBM POWER6<br />2 cores * 4 threads/core (May 2007)<br />Tilera 64 cores<br />Dan Reed, Microsoft<br />11<br />
    12. 12. Marching towards manycore<br />Intel’s 80 core prototype<br />2-D mesh interconnect<br />62 W power<br />Tilera 64 core system<br />8x8 grid of cores<br />5 MB coherent cache<br />4 DDR2 controllers<br />2 10 GbE interfaces<br />IBM Cell<br />PowerPC and 8 cores<br />Dan Reed, Microsoft<br />12<br />
    13. 13. 1E+17<br />multi-Petaflop<br />Petaflop<br />Blue Gene/L<br />1E+14<br />Thunder<br />Red Storm<br />Earth<br />Blue Pacific<br />ASCI White, ASCI Q<br />SX-5<br />ASCI Red Option<br />ASCI Red<br />T3E<br />SX-4<br />NWT<br />CP-PACS<br />1E+11<br />CM-5<br />Paragon<br />T3D<br />Delta<br />SX-3/44<br />Doubling time = 1.5 yr.<br />i860 (MPPs)<br />VP2600/10<br />SX-2<br />CRAY-2<br />Y-MP8<br />S-810/20<br />X-MP4<br />Peak Speed (flops)<br />Cyber 205<br />X-MP2 (parallel vectors)<br />1E+8<br />CRAY-1<br />CDC STAR-100 (vectors)<br />CDC 7600<br />ILLIAC IV<br />CDC 6600 (ICs)<br />IBM Stretch<br />1E+5<br />IBM 7090 (transistors)<br />IBM 704<br />IBM 701<br />UNIVAC<br />ENIAC (vacuum tubes)<br />1E+2<br />1940<br />1950<br />1960<br />1970<br />1980<br />1990<br />2000<br />2010<br />Year Introduced<br />The evolution of the fastest supercomputer<br />Argonne<br />My laptop<br />
    14. 14. The Argonne IBM BG/P<br />
    15. 15. www.top500.org<br />1<br />2<br />3-4<br />&gt;128K<br />
    16. 16. Simulation of the human arterial tree on the TeraGrid<br />G. Karniadakis et al.<br />
    17. 17.
    18. 18. Storage costs<br />(PC Magazine, Oct 2, 2007)<br />
    19. 19. Informationbig bang<br />All informationper year<br />100Exabytes<br />Uniqueinformationper year<br />All human documentsproduced last 40,000 years(to 1997)<br />12 Exabytes<br /> 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009<br />© Stuart Card, basedon Lesk, Berkeley SIMS, Landauer, EMC<br />
    20. 20. Growth of Genbank(1982-2005)<br />Broad Institute<br />
    21. 21. More data does not always mean more knowledge<br />Folker Meyer, Genome Sequencing vs. Moore’s Law: Cyber Challenges for the Next Decade, CTWatch, August 2006.<br />
    22. 22. The Red Queen’s race<br /> &quot;Well, in our country,&quot; said Alice … &quot;you&apos;d generally get to somewhere else — if you run very fast for a long time, as we&apos;ve been doing.”<br /> &quot;A slow sort of country!&quot; said the Queen. &quot;Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!&quot;<br />
    23. 23. Computing ondemand<br />Public PUMA knowledge base<br />Information about proteins analyzed against ~2 million gene sequences<br />Back officeanalysis on Grid<br />Millions of BLAST, BLOCKS, etc., onOSG and TeraGrid<br />Natalia Maltsev et al.<br />
    24. 24.
    25. 25.
    26. 26. 1.6 Tbps<br />Cost perGigabit-Mile<br />320 Gbps<br />Moore’sLaw<br />2.5 Gbps<br />50 Mbps<br />1984<br />1994<br />1998<br />2000<br />1993<br />1998<br />2002<br />Opticalnetworkingbreakthrough!<br />Revolution<br />Capacity increase and new economics<br />Nortel<br />
    27. 27. Optical switches<br />Lucent<br />
    28. 28. Empiricism<br />Theory<br />Simulation<br />Data<br />New ways of knowing<br />300 BCE<br />1700<br />1950<br />1990<br />Enhanced by the power of collaboration<br />
    29. 29.
    30. 30. Quantitative medicine is the key to reducing healthcare costs and improving healthcare outcomes<br />Patients with same diagnosis<br />
    31. 31. Quantitative medicine is the key to reducing healthcare costs and improving healthcare outcomes<br />Non-responderstoxic responders<br />Non-toxic responders<br />Patients with same diagnosis<br />Misdiagnosed<br />
    32. 32. Leukemia and Lymphoma<br />After Mara Aspinall, GenzymeGenetics; Felix W. Frueh, FDA <br />
    33. 33. Leukemia and Lymphoma<br />After Mara Aspinall, GenzymeGenetics; Felix W. Frueh, FDA <br />
    34. 34. Currently, 17% of Burkitt&apos;s Lymphoma are incorrectly diagnosed as Diffuse Large B Cell Lymphoma<br />Classic<br />Burkitt’sLymphoma<br />Atypical<br />Burkitt’sLymphoma<br />Diffuse Large<br />B Cell Lymphoma<br />Louis Staudt, National Cancer Institute<br />
    35. 35. Classic Burkitt Lymphoma Cure Rate<br />Diagnosis<br />Atypical Burkitt Lymphoma Cure Rate<br />Treatment<br />Patients’ Actual Disease<br />Burkitt’sLymphoma<br />60 %<br />80 %<br />Intensive chemotherapy<br />Burkitt’sLymphoma<br />Diffuse Large B Cell Lymphoma<br />0 %<br />15 %<br />CHOPregimen<br />Burkitt’sLymphoma<br />
    36. 36. Survival estimates for patients with Burkitt&apos;s Lymphoma<br />Best treatment for Burkitt’s Lymphoma<br />Best treatment for Diffuse Large B Cell Lymphoma<br />Dave et al, NEJM, June 8, 2006.<br />
    37. 37. Burkitt’s<br />Lymphoma<br />Diffuse Large<br /> B-cell Lymphoma<br />Classic Atypical<br />Louis Staudt, National Cancer Institute<br />
    38. 38. Imaging biomarkers: Diffusion Tensor Imaging and brain injury<br />Kraus et al., Brain (2007), 130, 2508-2519<br />
    39. 39. Enabling quantitative medicine<br />Collect a lot of patient data<br />Analyze data to infer effective treatments<br />Identify personalized treatment plans<br />Clinical practice<br />Basic research<br />Clinical trials<br />
    40. 40. Challenges<br />Increasing volumes of data, types of data: genomics, blood proteins, imaging, …<br />New science and treatments are hidden in the data, not the biology (biomarkers)<br />Too much for the individual physician or researcher to absorb<br /> … have to pay attention to cognitive support … computer-based tools and systems that offer clinicians and patients assistance for thinking about and solving problems related to specific instances of health care.<br />NRC Report on Computational Technology for Effective Health Care: Immediate Steps and Strategic Directions, 2009<br />
    41. 41. Bridging silos to enable quantitative medicine<br />Basic research<br />ongoing investigative studies<br />Outcomes, tissue bank<br />screening tests<br />pathways<br /> library<br />Clinical practice<br />Clinical trials<br />trial subjects, outcomes<br />
    42. 42. Addressing urban health needs<br />
    43. 43. Important characteristics<br />We must integrate systems that may not have worked together before<br />These are human systems, with differing goals, incentives, capabilities<br />All components are dynamic—change is the norm, not the exception<br />Processes are evolving rapidly too<br />We are not building something simple like a bridge or an airline reservation system<br />
    44. 44. Healthcare is acomplex adaptive system<br />A complex adaptive system is a collection of individual agents that have the freedom to act in ways that are not always predictable and whose actions are interconnected such that one agent’s actions changes the context for other agents.<br />Crossing the Quality Chasm, IOM, 2001; pp 312-13<br /><ul><li>Non-linear and dynamic
    45. 45. Agents are independent and intelligent
    46. 46. Goals and behaviors often in conflict
    47. 47. Self-organization through adaptation and learning
    48. 48. No single point(s) of control
    49. 49. Hierarchical decomp-osition has limited value</li></li></ul><li>We need to function in the zone of complexity<br />Low<br />Chaos<br />Zone ofcomplexity<br />Agreement<br />about<br />outcomes<br />Plan and control<br />High<br />Low<br />High<br />Certainty about outcomes<br />Ralph Stacey, Complexity and Creativity in Organizations, 1996<br />
    50. 50. We need to function in the zone of complexity<br />Low<br />Chaos<br />Agreement<br />about<br />outcomes<br />Plan and control<br />High<br />Low<br />High<br />Certainty about outcomes<br />Ralph Stacey, Complexity and Creativity in Organizations, 1996<br />
    51. 51. We call these groupingsvirtual organizations (VOs)<br />A set of individuals and/or institutions engaged in the controlled sharing of resources in pursuit of a common goal <br /> But U.S. health system is marked by fragmented and inefficient VOs with insufficient mechanisms for controlled sharing<br /> Healthcare = dynamic, overlapping VOs, linking<br />Patient – primary care<br />Sub-specialist – hospital<br />Pharmacy – laboratory<br />Insurer – …<br /> I advocate … a model of virtual integration rather than true vertical integration … G. Halvorson, CEO Kaiser<br />
    52. 52. The Grid paradigm<br />Principles and mechanisms for dynamic VOs<br />Leverage service oriented architecture (SOA)<br />Loose coupling of data and services<br />Open software,architecture<br />Engineering<br />Biomedicine<br />Computer science<br />Physics<br />Healthcare<br />Astronomy<br />Biology<br />1995 2000 2005 2010<br />
    53. 53. The Grid paradigm and healthcare information integration<br />[Grid architecture joint work with Carl Kesselman, Steve Tuecke, Stephan Erberich, and others] <br />Manage who can do what<br />Make data usable and useful<br />Platform services<br />Name data and move it around<br />Make data accessible over the network<br />Data sources<br />Radiology<br />Medical records<br />Pathology<br />Genomics<br />Labs<br />RHIO<br />
    54. 54. The Grid paradigm and healthcare information integration<br />Enhance user cognitive processes<br />Security and policy<br />Incorporate into business processes<br />Transform data into knowledge<br />Integration<br />Platform services<br />Management<br />Publication<br />Data sources<br />Radiology<br />Medical records<br />Pathology<br />Genomics<br />Labs<br />RHIO<br />
    55. 55. The Grid paradigm and healthcare information integration<br />Cognitive support<br />Security and policy<br />Valueservices<br />Applications<br />Analysis<br />Integration<br />Platform services<br />Management<br />Publication<br />Data sources<br />Radiology<br />Medical records<br />Pathology<br />Genomics<br />Labs<br />RHIO<br />
    56. 56. We partition the multi-faceted interoperability problem<br />Process interoperability<br />Integrate work across healthcare enterprise<br />Data interoperability<br />Syntactic: move structured data among system elements<br />Semantic: use information across system elements<br />Systems interoperability<br />Communicate securely, reliably among system elements<br />Applications<br />Analysis<br />Integration<br />Management<br />Publication<br />
    57. 57. Publication:Make information accessible<br />Make data available in a remotely accessible, reusable manner<br />Leave mediation for integration layer<br />Gateway from local policy/protocol into wide area mechanisms (transport, security, …)<br />
    58. 58. Imaging clinical trials<br />NeuroblastomaCancerFoundation<br />Childrens Oncology Group<br />
    59. 59. Stephan Erberich,<br />Carl Kesselman, et al.<br />
    60. 60. As of Oct19, 2008:<br />122 participants<br />105 services<br />70 data<br />35 analytical <br />
    61. 61. Data movement in clinical trials<br />(Center for Health Informatics)<br />
    62. 62. Community public health:Digital retinopathy screening network<br />(Center for Health Informatics)<br />
    63. 63. Integration:Making data usable and useful<br />?<br />Adaptive approach<br />100%<br />Degree of communication<br />Loosely coupled approach<br />Rigid standards-based approach<br />0%<br />0% 100%<br /> Degree of prior syntactic and semantic agreement<br />
    64. 64. Integration via mediation<br /><ul><li>Map between models
    65. 65. Scoped to domain use
    66. 66. Multiple concurrent use
    67. 67. Bottom up mediation
    68. 68. between standards and versions
    69. 69. between local versions
    70. 70. in absence of agreement</li></ul>Global Data Model<br />Query reformulation<br />Query in union of exported source schema<br />Query optimization<br />Distributed query execution<br />Query execution engine<br />Wrapper<br />Wrapper<br />Query in the <br />sourceschema<br />Alon Halevy, 2000<br />
    71. 71. Analytics:Transform data into knowledge<br /> “The overwhelming success of genetic and genomic research efforts has created an enormous backlog of data with the potential to improve the quality of patient care and cost effectiveness of treatment.”<br />— US Presidential Council of Advisors on Science and Technology, Personalized Medicine Themes, 2008<br />
    72. 72. Published<br />The imagepyramid<br />Enrolled/ evaluated<br />Eligible patients<br />Created<br />Michael Vannier<br />
    73. 73. Query and retrieve microarray data from a caArray data service:cagridnode.c2b2.columbia.edu:8080/wsrf/services/cagrid/CaArrayScrub<br />Normalize microarray data using GenePattern analytical servicenode255.broad.mit.edu:6060/wsrf/services/cagrid/PreprocessDatasetMAGEService<br />Hierarchical clustering using geWorkbench analytical service: cagridnode.c2b2.columbia.edu:8080/wsrf/services/cagrid/HierarchicalClusteringMage<br />Microarray clustering using Taverna<br />Workflow in/output<br />caGrid services<br />“Shim” services<br />others<br />Wei Tan et al.<br />
    74. 74. Many many tasks:Identifying potential drug targets<br /> 2M+ ligands<br /> Protein xtarget(s) <br />Benoit Roux et al.<br />
    75. 75. 6 GB<br />2M structures<br />(6 GB)<br />~4M x 60s x 1 cpu<br />~60K cpu-hrs<br />FRED<br />DOCK6<br />Select best ~5K<br />Select best ~5K<br />~10K x 20m x 1 cpu<br />~3K cpu-hrs<br />Amber<br />Select best ~500<br />~500 x 10hr x 100 cpu<br />~500K cpu-hrs<br />GCMC<br />ZINC<br />3-D<br />structures<br />Manually prep<br />DOCK6 rec file<br />Manually prep<br />FRED rec file<br />NAB scriptparameters<br />(defines flexible<br />residues, <br />#MDsteps)<br />NAB<br />Script<br />Template<br />DOCK6<br />Receptor<br />(1 per protein:<br />defines pocket<br />to bind to)<br />FRED<br />Receptor<br />(1 per protein:<br />defines pocket<br />to bind to)<br />PDB<br />protein<br />descriptions<br />1 protein<br />(1MB)<br />BuildNABScript<br />Amber prep:<br />2. AmberizeReceptor<br />4. perl: gen nabscript<br />NAB<br />Script<br />start<br />Amber Score:<br />1. AmberizeLigand<br />3. AmberizeComplex<br />5. RunNABScript<br />For 1 target:<br />4 million tasks500,000 cpu-hrs<br />(50 cpu-years)<br />end<br />report<br />ligands<br />complexes<br />
    76. 76. DOCK on BG/P: ~1M tasks on 118,000 CPUs<br />CPU cores: 118784<br />Tasks: 934803<br />Elapsed time: 7257 sec<br />Compute time: 21.43 CPU years<br />Average task time: 667 sec<br />Relative Efficiency: 99.7% (from 16 to 32 racks)<br />Utilization: <br />Sustained: 99.6%<br />Overall: 78.3%<br />Time (secs)<br />Ioan Raicu et al.<br />
    77. 77. The health care information technology chasm<br /> Health care IT [is] rarely used to provide clinicians with evidence-based decision support and feedback; to support data-driven process improvement; or to link clinical care and research.<br />Computational Technology for Effective Health Care, NRC, 2009<br />
    78. 78. Six research challenges for information technology and healthcare<br />Patient-centered cognitive support<br />Modeling—an individualized virtual patient<br />Automation—integrated use, adaptivity<br />Data sharing and collaboration<br />Data management at scale<br />Automated full capture of physician-patient interactions<br />Computational Technology for Effective Health Care, NRC, 2009<br />
    79. 79. Six research challenges for information technology and healthcare<br />Patient-centered cognitive support<br />Modeling—an individualized virtual patient<br />Automation—integrated use, adaptivity<br />Data sharing and collaboration<br />Data management at scale<br />Automated full capture of physician-patient interactions<br />Computational Technology for Effective Health Care, NRC, 2009<br />
    80. 80. Functioning in the zone of complexity<br />Low<br />Chaos<br />Agreement<br />about<br />outcomes<br />Plan and control<br />High<br />Low<br />High<br />Certainty about outcomes<br />Ralph Stacey, Complexity and Creativity in Organizations, 1996<br />
    81. 81. The Grid paradigm and healthcare information integration<br />Cognitive support<br />Security and policy<br />Valueservices<br />Applications<br />Analysis<br />Integration<br />Platform services<br />Management<br />Publication<br />Data sources<br />Radiology<br />Medical records<br />Pathology<br />Genomics<br />Labs<br />RHIO<br />
    82. 82. “People tend to overestimate the short-term impact of change, and underestimate the long-term impact.”<br /> — Roy Amara<br /> “The computer revolution hasn’t happened yet.” <br /> — Alan Kay, 1997<br />
    83. 83. Thank you!<br />Computation Institutewww.ci.uchicago.edu<br />

    ×