![Markov Decision ProcessesActions have probabilistic effectsTreatments sometimes workTesting can have effectsThe probabilities depend on the patient’s state and the actions Actions have costsThe patient’s state has an immediate valueQuality of lifeM = <S, A, Pr, R>, Pr: SxAxS [0,1]](https://image.slidesharecdn.com/pptx-preventing-sepsis-artificial-intelligence-knowledge3211/85/pptx-Preventing-Sepsis-Artificial-Intelligence-Knowledge-21-320.jpg)
pptx - Preventing Sepsis: Artificial Intelligence, Knowledge ...
- 1. Preventing Sepsis: Artificial Intelligence, Knowledge Discovery, & VisualizationPhillip Chang, MD (Dept of Surgery) Judy Goldsmith, PhD (Dept of Computer Science)Remco Chang, PhD (UNC-Charlotte Visualization Center)
- 2. NIH Challenge GrantThis application addresses broad Challenge Area (10) Information Technology for Processing Health Care Data Topic, 10-LM-102*: Advanced decision support for complex clinical decisions
- 3. Clinical Problem: sepsisDefinition: serious medical condition characterized by a whole-body inflammatory state (called a systemic inflammatory response syndrome or SIRS) and the presence of a known or suspected infectionTop 10 causes of death in the USKills more than 200,000 per year in the US (more than breast & lung cancer combined)
- 4. Cost of severe sepsisEstimated cases per year in US: 751,000Estimated cost per case: $22,100Estimated total cost per year: $16.7 billionMortality (in this series): 28%Projected increase 1.5% per annumAngus et al. Epidemiology of severe sepsis in the United States: Analysis of incidence, outcome, and associated costs of care. Critical Care Medicine. July, 2001
- 5. SIRSTemperature < 36° C or > 38° C Heart Rate > 90 bpmRespiratory Rate > 20 breaths/minor PaCO2 < 32 mmHg White Blood Cell Count > 12,000 or < 4,000 cells/mm3; or > 10% bands
- 8. 2008 versionMortality remains 35-60%
- 9. What’s the problem?Early recognitionBiomarkers?Equivalent of troponin-I for sepsisAlert systems?
- 10. BiomarkersNot a single marker exist, yet….
- 11. Alert SystemsTrue alertsNeither sensitive nor specificCannot find “sweet-spot”We’re working on one now….Other forms are “early recognition”
- 13. What about Bob?
- 14. Our premiseRetrospective chart review often yields time frame when one feels early intervention could have changed outcomeClinical “hunch” that something “bad” might happen which demands more attentionWhat if we could predict sepsis before sepsis criteria were met?
- 15. Our goal
- 16. How do we do this?Data MiningArtificial IntelligenceVisualization (computer-human interface)
- 17. Data! Data! Data! Heartrate??????TemperaturePaCO2Respiratory Rate White Blood Cell Count
- 18. Marriage of computer science & medicineData miningidentify previously undiscovered patterns and correlationsChanges in vital signsRate of change of the vitals signsPerhaps correlations of seemingly unrelated eventsRecently found that prior to significant hemodynamic compromise, the variation in heart rate actually decreases in mice
- 19. Marriage of computer science & medicineDecision makingIncreased monitoring of vitals?More tests? (Which ones?)Antibiotics?Exploratory surgery?None of the above?What drives decisions?Costs, benefitsLikelihood of benefits
- 20. Marriage of computer science & medicineArtificial IntelligenceModel knowledge (from data mining) into partially observable Markov decision process (POMDP)
- 21. Markov Decision ProcessesActions have probabilistic effectsTreatments sometimes workTesting can have effectsThe probabilities depend on the patient’s state and the actions Actions have costsThe patient’s state has an immediate valueQuality of lifeM = <S, A, Pr, R>, Pr: SxAxS [0,1]
- 22. Decision-Theoretic Planning“Plans” are policies: Given the patient’s history, the insurance plan (establishes costs)probabilities of effectsOptimize long term expected outcomes(That’s a lot of possibilities, even for computers!)(π: S A)
- 23. Partially Observable MDPsThe patient’s state is not fully observableThis makes planning harderPut probabilities on unobserved variablesReason over possible states as well as possible futures(π: Histories A)Optimality is no longer feasible Don’t despair! Satisficing policies are possible.
- 24. AI SummaryUse data mining, machine learning to find patterns and predictorsBuild POMDP model Find policy that considers long-term expected costsGet alerts when sepsis is likely, suggested tests or treatments that are cost- and outcome-effective
- 25. NASA used it….To reduce “cognitive load”
- 30. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 31. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 32. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 33. Values of VisualizationPresentationAnalysis>>Slide courtesy of Dr. Pat Hanrahan, Stanford
- 34. Values of VisualizationPresentationAnalysis3.14286 3.140845>>Slide courtesy of Dr. Pat Hanrahan, Stanford
- 35. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 36. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 37. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 38. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 39. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 40. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 41. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 42. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 43. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 44. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 45. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 46. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 47. Values of VisualizationPresentationAnalysisSlide courtesy of Dr. Pat Hanrahan, Stanford
- 48. Values of VisualizationPresentationAnalysis?Slide courtesy of Dr. Pat Hanrahan, Stanford
- 49. Using Visualizations To Solve Real-World Problems…
- 50. Using Visualizations To Solve Real-World Problems…WhoWhereWhatEvidenceBoxOriginal DataWhen
- 51. Using Visualizations To Solve Real-World Problems…This group’s attacks are not bounded by geo-locations but instead, religious beliefs. Its attack patterns changed with its developments.
- 52. Visualization conceptIt’s your consigliere – always there, in the background
- 53. Visualizing SepsisChallengesConnecting to Data Mining and AI componentsDoctors don’t sit in front of a computer all the time…
- 54. ValidationModel will need to be built on retrospective dataValidated on real-time prospective dataClinical trial?
- 55. Leap of faith?
Editor's Notes
- #27: 2008 – 455 billion2009 – proposed 533 billion
- #28: 2008 – 455 billion2009 – proposed 533 billion
- #29: 2008 – 4552009 – proposed 533
- #30: I recently finished reading a wonderful book by Steven Johnson entitled The Ghost Map: The Story of London’s Most Terrifying Epidemic – and How It Changed Science, Cities, and the Modern World. In the summer of 1854 cholera swept through a section of London with unprecedented intensity. At the time, the cause of cholera was unknown and rapidly growing modern cities such as London, with dense populations packed into small areas, were rich breeding grounds for this disease. Most of those who concerned themselves with disease and its cure held tightly to the miasma theory that cholera spread through the air and was associated with the bad smells and the unclean urban environments that produced them. In fact, cholera is a bacterium, which was spreading through the water supply. This book tells the story much as a journalist who witnessed it firsthand would do, but a journalist who had the advantage of hindsight informed by knowledge of modern medicine.Several people of the time play important roles in this story – none more than John Snow, a medical doctor and research scientist. The ghost map refers to a map that he drew by hand during the process of his investigations, which could clearly demonstrate to anyone with open eyes that the source of the outbreak was the Broad Street well. Despite the evidence that this map displayed, however, the miasma theory of cholera transmission prevailed for several years after the epidemic. Eventually, due largely to the tenacious efforts of John Snow and an unlikely supporter, Reverend Henry Whitehead, the evidence won out and steps were taken to eliminate the conditions in which cholera could spread.
- #34: 22/7: 3.14286223/71: 3.140845
- #35: 22/7: 3.14286223/71: 3.140845