PhD Defense
Upcoming SlideShare
Loading in...5
×

Like this? Share it with your network

Share
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
472
On Slideshare
472
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
16
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis Conclusions On the Intelligent Management of Sepsis in the Intensive Care Unit Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) Barcelona January 28, 2013Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 2. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 3. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 4. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 5. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroductionVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 6. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroductionIntroduction Sepsis is a clinical syndrome defined by the presence of infection and Systemic Inflammatory Response Syndrome (SIRS).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 7. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroductionIntroduction Sepsis is a clinical syndrome defined by the presence of infection and Systemic Inflammatory Response Syndrome (SIRS). This can lead to severe sepsis or to septic shock (severe sepsis with hypotension refractory to fluid administration) and multi-organ failure.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 8. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroductionIntroduction Sepsis is a clinical syndrome defined by the presence of infection and Systemic Inflammatory Response Syndrome (SIRS). This can lead to severe sepsis or to septic shock (severe sepsis with hypotension refractory to fluid administration) and multi-organ failure. In western countries, septic patients account for as much as 25% of ICU bed utilization and the pathology occurs in 1% - 2% of all hospitalizations.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 9. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroductionIntroduction Sepsis is a clinical syndrome defined by the presence of infection and Systemic Inflammatory Response Syndrome (SIRS). This can lead to severe sepsis or to septic shock (severe sepsis with hypotension refractory to fluid administration) and multi-organ failure. In western countries, septic patients account for as much as 25% of ICU bed utilization and the pathology occurs in 1% - 2% of all hospitalizations. The mortality rates of sepsis range from 12.8% for sepsis and 20.7% for severe sepsis, and up to 45.7% for septic shock.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 10. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroduction The medical management of sepsis and the study of its prognosis and outcome is a relevant medical research challenge.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 11. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroduction The medical management of sepsis and the study of its prognosis and outcome is a relevant medical research challenge. Provided that such methods are to be used in a clinical environment (ICU), it requires prediction methods that are robust, accurate and readily interpretable.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 12. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIntroduction The medical management of sepsis and the study of its prognosis and outcome is a relevant medical research challenge. Provided that such methods are to be used in a clinical environment (ICU), it requires prediction methods that are robust, accurate and readily interpretable.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 13. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsThesis ObjectivesThesis Objectives Improve our knowledge about the incidence of Sepsis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 14. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsThesis ObjectivesThesis Objectives Improve our knowledge about the incidence of Sepsis. Improve our understanding about Sepsis and its protective factors.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 15. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsThesis ObjectivesThesis Objectives Improve our knowledge about the incidence of Sepsis. Improve our understanding about Sepsis and its protective factors. Study of the evolution of Sepsis into more critical states with respect to several management/measurement variables.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 16. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsThesis ObjectivesThesis Objectives Improve our knowledge about the incidence of Sepsis. Improve our understanding about Sepsis and its protective factors. Study of the evolution of Sepsis into more critical states with respect to several management/measurement variables. Develop a system that could provide prognostic indicators of mortality (RoD) that can be used in the ICU (acc., sens, spec), if possible, at the onset of the pathology.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 17. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 18. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsDatasetAvailable Datasets A prospective observational cohort study of adult patients with severe sepsis was conducted at the Critical Care Department of the Vall d’ Hebron University Hospital (Barcelona, Spain).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 19. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsDatasetAvailable Datasets A prospective observational cohort study of adult patients with severe sepsis was conducted at the Critical Care Department of the Vall d’ Hebron University Hospital (Barcelona, Spain). Data from 750 and 354 patients with severe sepsis was collected in this ICU between June, 2007 and December, 2010.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 20. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsDatasetAvailable Datasets A prospective observational cohort study of adult patients with severe sepsis was conducted at the Critical Care Department of the Vall d’ Hebron University Hospital (Barcelona, Spain). Data from 750 and 354 patients with severe sepsis was collected in this ICU between June, 2007 and December, 2010. 55% of cases correspond to ‘medical’ sepsis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 21. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsDatasetAvailable Datasets A prospective observational cohort study of adult patients with severe sepsis was conducted at the Critical Care Department of the Vall d’ Hebron University Hospital (Barcelona, Spain). Data from 750 and 354 patients with severe sepsis was collected in this ICU between June, 2007 and December, 2010. 55% of cases correspond to ‘medical’ sepsis. The mean age of the patients in the database was 57.08 (with standard deviation ±16.65) years.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 22. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsDatasetAvailable Datasets A prospective observational cohort study of adult patients with severe sepsis was conducted at the Critical Care Department of the Vall d’ Hebron University Hospital (Barcelona, Spain). Data from 750 and 354 patients with severe sepsis was collected in this ICU between June, 2007 and December, 2010. 55% of cases correspond to ‘medical’ sepsis. The mean age of the patients in the database was 57.08 (with standard deviation ±16.65) years. 40% of patients were female.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 23. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAvailable DataAvailable Attributes The collected data show the worst values for all variables during the first 24 hours of evolution of Severe Sepsis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 24. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAvailable DataAvailable Attributes The collected data show the worst values for all variables during the first 24 hours of evolution of Severe Sepsis. Organ dysfunction was evaluated by means of the SOFA score system, which objectively measures organ dysfunction for 6 organs/systems. Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14) Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86) Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs 1.68 (1.09) Hepatic (HEPA) 0.48 (0.92) Failure Organs 1.51 (1.02) Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32)Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 25. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAvailable DataAvailable Attributes The collected data show the worst values for all variables during the first 24 hours of evolution of Severe Sepsis. Organ dysfunction was evaluated by means of the SOFA score system, which objectively measures organ dysfunction for 6 organs/systems. Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14) Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86) Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs 1.68 (1.09) Hepatic (HEPA) 0.48 (0.92) Failure Organs 1.51 (1.02) Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32) Severity was evaluated by means of the APACHE II score, which was 23.03 ± 9.62 for the population under study.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 26. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAvailable DataAvailable Attributes The collected data show the worst values for all variables during the first 24 hours of evolution of Severe Sepsis. Organ dysfunction was evaluated by means of the SOFA score system, which objectively measures organ dysfunction for 6 organs/systems. Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14) Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86) Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs 1.68 (1.09) Hepatic (HEPA) 0.48 (0.92) Failure Organs 1.51 (1.02) Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32) Severity was evaluated by means of the APACHE II score, which was 23.03 ± 9.62 for the population under study. The mortality rate intra-ICU for our study population was 26.32%.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 27. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAvailable DataAvailable Attributes List of attributes used in this study: Age Haemocultures 6h Dysfunctional Organs for Gender SOFA 1-2 Antibiotics 6h Sepsis Focus Dysfunctional Organs for Volume 6h SOFA 3-4 O2 Central Venous Saturation Germ Class Total Number of 6h Polimicrobial Infection Dysfunctional Organs Haematocrit 6h Base Pathology Mechanical Ventilation Transfusions 6h Cardiovascular SOFA Score Oxygenation Index Dobutamine 6h Respiratory SOFA Score PaO2 /FiO2 Surviving Sepsis Campaign CNS SOFA score Vasoactive Drugs Bundles 24h Hepatic SOFA Score Platelet Count Glycaemia 24h Renal SOFA Score APACHE II Score PPlateau Haematologic SOFA Score Surviving Sepsis Campaign Worst Lactate Total SOFA Score Bundles 6h O2 Central Venous SaturationVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 28. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 29. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsQuantitative Analysis of the Pathophysiology of Sepsis Reference Objective Model Variables Rackow et al. 1991 Severity ODE B. Flow MAP Art. Res React. Hyperemia Kimberly et al. 2000 Shock Autonomic HR Coupling BP Ross et al. 1998 Inflam. ODE Pathogen Shock ANN (RBF) Cell Damage Immuno. Resp. Othman et al. 2003 Shock Gastric Impedance and 2004 MucosaVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 30. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsQuantitative Analysis of the Prognosis of Sepsis (MedicalPractice) Reference Objective Model Variables Knaus et al. 1985 Eval. LR SOFA MODS APACHE II LODS Adler et al. 2008 Sepsis LR SAPS 3 Paetz et al. 2001 Shock RBF SBP Thrombocit. Lact. Savkin et al. 2010 Sepsis SVM Culture RR, HR Lact. White cell Brause et al. 2002 Sepsis HMM Sep. States Full Obs.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 31. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 32. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAI Methods AppliedVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 33. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsAI Methods Applied Graphical models from Algebraic Statistics chapter 6. CART chapter 6. Logistic regression over latent factors chapter 7. Shrinkage methods chapter 8. [Hastie] Support Vector Machines chapter 8 [Sch¨lkopf, Christianini]. o Generative kernels chapter 8. Simplified Fisher kernel. Quotient Basis Kernel. Kernels based on the JS metric. Relevance Vector Machines chapter 8 [Tipping, Fletcher].Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 34. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsShrinkage methods M Linear Regression: minw Lλ (w, y, x) = i=1 (yi − g (xi ))2 ,Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 35. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsShrinkage methods M Linear Regression: minw Lλ (w, y, x) = i=1 (yi − g (xi ))2 , Ridge regression: M minw Lλ (w, y, x) = minw λ w 2 + i=1 (yi − g (xi ))2 ,Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 36. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsShrinkage methods M Linear Regression: minw Lλ (w, y, x) = i=1 (yi − g (xi ))2 , Ridge regression: M minw Lλ (w, y, x) = minw λ w 2 + i=1 (yi − g (xi ))2 , Lasso (L1 loss function): M minw Lλ (w, y, x) = minw λ|w|1 + i=1 (yi − g (xi ))2 ,Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 37. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsShrinkage methods M Linear Regression: minw Lλ (w, y, x) = i=1 (yi − g (xi ))2 , Ridge regression: M minw Lλ (w, y, x) = minw λ w 2 + i=1 (yi − g (xi ))2 , Lasso (L1 loss function): M minw Lλ (w, y, x) = minw λ|w|1 + i=1 (yi − g (xi ))2 , Relevance vector machine (SVM with priors) [Tipping, Fletcher].Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 38. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsWhy generative kernels? Selection of kernel function for a given problem is not trivial.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 39. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsWhy generative kernels? Selection of kernel function for a given problem is not trivial. One normally must have good insight about the problem at hand.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 40. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsWhy generative kernels? Selection of kernel function for a given problem is not trivial. One normally must have good insight about the problem at hand. Mapping over higher dimensions simplifies the problem but computational cost grows with ∼ d 3 .Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 41. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsWhy generative kernels? Selection of kernel function for a given problem is not trivial. One normally must have good insight about the problem at hand. Mapping over higher dimensions simplifies the problem but computational cost grows with ∼ d 3 . Solution: exploit the statistical structure of the data to build the kernel!Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 42. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsWhy generative kernels? Selection of kernel function for a given problem is not trivial. One normally must have good insight about the problem at hand. Mapping over higher dimensions simplifies the problem but computational cost grows with ∼ d 3 . Solution: exploit the statistical structure of the data to build the kernel! Requirement: pdf must be a regular exponential family. This requirement is fulfilled by the dataset of this Ph.D.!Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 43. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsGenerative kernels We propose three generative approaches:Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 44. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsGenerative kernels We propose three generative approaches: exploit the algebraic structure of the dataset (Quotient Basis).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 45. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsGenerative kernels We propose three generative approaches: exploit the algebraic structure of the dataset (Quotient Basis). exploit the momentum generation properties of regular exponential families (i.e. the second derivative of the log-Laplace function G corresponds to the covariance) to calculate the Fisher kernel.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 46. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsGenerative kernels We propose three generative approaches: exploit the algebraic structure of the dataset (Quotient Basis). exploit the momentum generation properties of regular exponential families (i.e. the second derivative of the log-Laplace function G corresponds to the covariance) to calculate the Fisher kernel. use the dual of the log-Laplace function, which corresponds to the negative entropy, and a metric (Jensen-Shannon) to build the kernels through application of the properties for pd and nd functions (i.e. centering, inversion and exponentiation).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 47. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsQuotient Basis Kernel Let A be a set of n unique points A = {a1 , . . . , an } and τ a term ordering. A Gr¨bner basis of A, G = g1 , . . . , gt , is a Gr¨bner basis o o of I (A). Therefore, the points in A can be presented as the set of solutions of   g1 (a) = 0  g2 (a) = 0    ··· gt (a) = 0 Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 48. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsQuotient Basis Kernel Let A, be a set of n × s unique support points A = {a1 , . . . , an } and τ a term ordering. A monomial basis of the set of polynomial functions over A is ESTτ = {x α : x α ∈ LT(g ) : g ∈ I (A) } / This means that ESTτ comprises the elements x α that are not divisible by any of the leading terms of the elements of the Gr¨bner o basis of I (A).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 49. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsQuotient Basis Kernel Let τ be a term ordering and let us consider an ordering over the support points A = {a1 , . . . , an }. We call design matrix (i.e. ESTτ evaluated in A) the following n × c matrix Z = [ESTτ ] A where c is the cardinality of ESTτ and n is the number of support points. The covariance of the design matrix of ESTτ , which is a kernel, is the QBK.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 50. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsSimplified Fisher Kernel Let P = (P|η ∈ N) be a regular exponential family with canonical sufficient statistic T . Then the log likelihood function takes the form l(η|T ) = n(η t T − G (η)) The score function is the gradient ∂l(η|T ) ∂ U(T , η) = = nT − G (η) ∂η ∂η So the simplified Fisher Kernel is: k(x, z) = U(Tx , η)U(Tz , η)tVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 51. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsKernels based on the Jensen Shannon MetricVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 52. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsKernels based on the Jensen Shannon Metric Let P = (P|η ∈ N) be a regular exponential family with canonical sufficient statistic T . Then the log likelihood function takes the form l(η|T ) = n(η t T − G (η)) This function accepts a convex - conjugate (Legendre Dual) of the form l(γ|T ) = n(γ t T − F (γ)) In our case, the dual F is the negative log-entropy function.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 53. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsKernels based on the Jensen Shannon Metric F (γ1 ) + F (γ2 ) γ1 + γ2 JS(γ1 , γ2 ) = −F . 2 2 Centred kernel: φ(x, y ) = JS(x, x0 ) + JS(y , x0 ) − JS(x, y ) − JS(x0 , x0 ).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 54. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsKernels based on the Jensen Shannon Metric F (γ1 ) + F (γ2 ) γ1 + γ2 JS(γ1 , γ2 ) = −F . 2 2 Centred kernel: φ(x, y ) = JS(x, x0 ) + JS(y , x0 ) − JS(x, y ) − JS(x0 , x0 ). Exponentiated kernel: φ(x, y ) = exp(−tJS(x, y )) ∀t > 0.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 55. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsKernels based on the Jensen Shannon Metric F (γ1 ) + F (γ2 ) γ1 + γ2 JS(γ1 , γ2 ) = −F . 2 2 Centred kernel: φ(x, y ) = JS(x, x0 ) + JS(y , x0 ) − JS(x, y ) − JS(x0 , x0 ). Exponentiated kernel: φ(x, y ) = exp(−tJS(x, y )) ∀t > 0. 1 Inverse kernel: φ(x, y ) = t+JS(x,y ) ∀t > 0.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 56. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 57. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRoute mapVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 58. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of SepsisSOFA Score and Sepsis Study based on the basal SOFA scale.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 59. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of SepsisSOFA Score and Sepsis Study based on the basal SOFA scale. A SOFA≥2 is demonstrative of MODS while a SOFA-CV > 2 is demonstrative of Septic Shock.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 60. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of SepsisSOFA Score and Sepsis Study based on the basal SOFA scale. A SOFA≥2 is demonstrative of MODS while a SOFA-CV > 2 is demonstrative of Septic Shock. By the definition of SOFA it is obvious that Severe Sepsis, Shock and MODS are dependent on each other (i.e. all imply organ dysfunction).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 61. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of SepsisBayes Network for Sepsis node X1 corresponds to the X1 X2 unobserved number of Severe Sepsis. node X2 corresponds to Septic Shock. node X3 corresponds to MODS. X3 X4 node X4 corresponds to ICU result.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 62. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of SepsisIncidence of Sepsis Bayes Network yields an incidence of 164 cases / 100000 hab (i.e. 164 vs 118 in our database). The incidence reported in Madrid is 141 cases / 100000 hab. and Castilla Le´n is 250 cases / 100000 hab. o The discrepancy between these figures lies mainly in the study design (haemocultures at admission vs. people being treated for Sepsis).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 63. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisStatins Statins is a common drug for the treatment of high cholesterol levels. Beyond their hypolipemic properties, they also exercise anti-inflammatory, immunomodulator and antioxidant actions. Statins modulate vasoreactivity in the coagulation system. Recent studies suggest that they present beneficial effects for infection prevention and treatment, impacting the ICU outcome. These results are still controversial in the medical community. In our dataset with 750 patients, 106 (i.e. 14.13%) received preadmission treatment with statins.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 64. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisHypothesis In our dataset with 750 patients, 106 (i.e. 14.13%) received preadmission treatment with statins. Do statins play a protective role in the prognosis of Sepsis? Does this role depend on the Sepsis continuum (MODS and Septic Shock)?Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 65. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisModels of Conditional Independence We aim to find the relation between the administration of statin drugs prior to ICU admission and the mortality rate in Severe Sepsis patients. Thus, we test the null hypothesis that the ICU outcome is independent of the preadmission use of statins for given APACHE II and SOFA scores. Ho : Ho : {X1 } ⊥ {X4 }|{X2 }, {X3 }. ⊥ (1) where {X1 } is the ICU outcome, {X4 } the preadmission use of statins, {X3 } the SOFA score, and {X2 } the APACHE II score.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 66. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisModels of Conditional Independence We have two options based on Algebraic Models to reject Ho : study the rank of the minors of our observation matrix (tedious for large datasets). Algebraic Interpolation (more fun): Assume that our observed dataset are the zeroes of polynomial set of equations (and an order on the variables).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 67. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisModels of Conditional Independence We have two options based on Algebraic Models to reject Ho : study the rank of the minors of our observation matrix (tedious for large datasets). Algebraic Interpolation (more fun): Assume that our observed dataset are the zeroes of polynomial set of equations (and an order on the variables). On a first level of abstraction there is a Polynomial Ideal associated to this set of equations (vanishing ideal).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 68. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisModels of Conditional Independence We have two options based on Algebraic Models to reject Ho : study the rank of the minors of our observation matrix (tedious for large datasets). Algebraic Interpolation (more fun): Assume that our observed dataset are the zeroes of polynomial set of equations (and an order on the variables). On a first level of abstraction there is a Polynomial Ideal associated to this set of equations (vanishing ideal). On a second level of abstraction, this ideal is generated by a finite basis (Hilbert basis theorem). This is a Gr¨bner basis. oVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 69. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisModels of Conditional Independence We have two options based on Algebraic Models to reject Ho : study the rank of the minors of our observation matrix (tedious for large datasets). Algebraic Interpolation (more fun): Assume that our observed dataset are the zeroes of polynomial set of equations (and an order on the variables). On a first level of abstraction there is a Polynomial Ideal associated to this set of equations (vanishing ideal). On a second level of abstraction, this ideal is generated by a finite basis (Hilbert basis theorem). This is a Gr¨bner basis. o On a third level of abstraction, the terms of the ordering that are not divided by the leading terms of the Gr¨bner basis are o defined as a Quotient Basis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 70. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisAlgebraic Interpolation The resulting products between terms of the Quotient Basis Statins SOFA APACHE II 1 1 1 show the 2 1 1 interaction/dependence 1 2 1 between terms. 2 2 1 1 1 2 By the Hammersley - Clifford 2 1 2 theorem (factorisation of the 1 2 2 terms of the Quotient Basis) 2 2 2 show that there is a Graphical Model associated to the vanishing ideal.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 71. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisAlgebraic Interpolation Ordering: x1 x2 x3 x1 x2 x1 x3 x1 x2 x1 x2 x3 x2 x3 1 Ideal I = 2 2 2 x3 − 3x3 + 2, x2 − 3x2 + 2, x1 − 3x1 + 2 . Gr¨bner G = o 2 − 3x + 2, x 2 − 3x + 2, x 2 − 3x + 2 . x3 3 2 2 1 1 Quotient Basis x3 x4 B = {1, x3 , x2 , x2 x3 , x1 , x1 x3 , x1 x2 , x1 x2 x3 }.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 72. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisAlgebraic Interpolation P(x4 ) = η7 x1 x2 x3 + η6 x1 x2 + η5 x1 x3 − η4 x2 x3 − η3 x1 + η2 x2 + η1 x3 + η0 Solving for {x1 , x2 , x3 } by substitution and also knowing that η0 = 1 − 7 ηi yields the interpolation polynomial i=1 P(x4 ) = −1/50x1 x2 x3 + 3/100x1 x2 + 9/100x1 x3 − 3/25x2 x3 − 21/100x1 + 27/100x2 + 8/25x3 + 7/25Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 73. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisAlgebraic Interpolation Statins SOFA APACHE II Result=1 Result=2 1 1 1 0.64 0.36 2 1 1 0.53 0.47 1 2 1 0.80 0.20 2 2 1 0.70 0.30 1 1 2 0.91 0.09 2 1 2 0.87 0.13 1 2 2 0.93 0.07 2 2 2 0.88 0.12Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 74. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisAlgebraic Interpolation Statins SOFA APACHE II Result=1 Result=2 1 1 1 0.64 0.36 2 1 1 0.53 0.47 1 2 1 0.80 0.20 2 2 1 0.70 0.30 1 1 2 0.91 0.09 2 1 2 0.87 0.13 1 2 2 0.93 0.07 2 2 2 0.88 0.12Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 75. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtection against SepsisAnalysis with Decision TreesVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 76. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationFactor AnalysisVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 77. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationFactor Analysis A Factor Analysis (FA) model concerns a Gaussian hidden variable model with d observed variables Xi and k hidden variables Yi . FA assumes (X , Y ) follows a joint multivariate normal distribution with positive definite covariance matrix. What is sought is a model of the form x − µ = Λ + .Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 78. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationFactor Analysis The FA model Fd,k is the family of multivariate normal distributions Nd (µ, Σ) on Rd whose mean vector µ is an arbitrary vector in Rd and whose covariance matrix Σ lies in the (non-convex) cone Fd,k = {Ω + ΛΛt ∈ Rd×d : Ω 0 diagonal, Λ ∈ Rd×k } = {Ω + Ψ ∈ Rd×d : Ω 0 diagonal, Ψ 0 symmetric, rank (Ψ) ≤ k}.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 79. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationFactor Analysis Factor 1: cardiovascular function, cardiovascular SOFA score and vasoactive drugs. Factor 2: haematologic function (haematologic SOFA score and platelet count). Factor 3: respiratory function, Respiratory SOFA score and PaO2 /FiO2 ratio. Factor 4: use of mechanical ventilation and PPlateau. Factor 5: 24h SSC bundles and glycaemic indices. Factor 6: micro-organism producing the Sepsis and whether this sepsis polimicrobial or not. Factor 7: renal function measured by the SOFA score and total SOFA score.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 80. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationFactor Analysis Factor 8: antibiotics and haemocultures during the first 6h of ICU stay. Factor 9: number of organs in dysfunction. Factor 10: hepatic function measured by the SOFA score. Factor 11: CNS function and number of organs in dysfunction. Factor 12: loci of Sepsis and poly-microbial. Factor 13: APACHE II score and worst lactate levels. Factor 14: Total number of organs in dysfunction.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 81. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationLogistic Regression Log-odd ratio of a Binomial distribution is p log = β0 + βX. (2) 1−p Where β0 is the intercept and β is vector of logistic regression coefficients estimated through ML with a generalised linear model.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 82. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationExperiment setting 10-Fold cross-validation. LR over Latent Factors (backward feature selection). Subset selection of the Original Variables (backward feature selection). Comparison with RoD formula based on the APACHE II score ROD ln 1−ROD = −3.517 + 0.146 · A + .Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 83. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationLogistic Regression over Latent Factors β Coeff MAX MIN Z-score Intercept 1.22 1.53 .87 7.11 F4 -0.54 -0.23 -0.86 -3.38 F10 -0.69 -0.38 -1.05 -4.26 F9 -0.51 -0.21 -0.81 -3.36 F13 -0.49 -0.24 -0.74 -3.80Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 84. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationLogistic Regression over Latent Factors β Coeff MAX MIN Z-score Intercept 4.20 3.11 5.29 7.56 APACHE II -0.08 -0.13 -0.04 -3.77 Worst Lact. -0.25 -0.38 -0.11 -3.63Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 85. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Prediction with a Latent Data RepresentationResults Method AUC Error Rate Sens. Spec. Dataset LR-FA 0.78 0.24 0.65 0.80 FA LR 0.75 0.30 0.64 0.72 LR APACHE II 0.70 0.28 0.82 0.55 N/AVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 86. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with RVM The model performance was evaluated by means of 10-Fold Cross-Validation.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 87. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with RVM The model performance was evaluated by means of 10-Fold Cross-Validation. The RVM yielded an accuracy of mortality prediction of 0.86; a prediction error of 0.18; a sensitivity of 0.67; and a specificity of 0.87.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 88. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with RVM The model performance was evaluated by means of 10-Fold Cross-Validation. The RVM yielded an accuracy of mortality prediction of 0.86; a prediction error of 0.18; a sensitivity of 0.67; and a specificity of 0.87. RVM selected the following attributes (corresponding to weights): Number of dysfunctional organs (w1 = −0.039) Mechanical Ventilation (w2 = −0.101) APACHE II (w3 = −0.337) Resuscitation Bundles (6h) (w4 = 0.037)Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 89. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed Data The coefficients corresponding to the rest of attributes were set to values close to zero as part of the training process. This reduces the number of attributes (34 to just 4) and improves its interpretability.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 90. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed Data The coefficients corresponding to the rest of attributes were set to values close to zero as part of the training process. This reduces the number of attributes (34 to just 4) and improves its interpretability. The negative weights (number of dysfunctional organs, mechanical ventilation, APACHE II) are related to a higher mortality risk.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 91. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed Data The coefficients corresponding to the rest of attributes were set to values close to zero as part of the training process. This reduces the number of attributes (34 to just 4) and improves its interpretability. The negative weights (number of dysfunctional organs, mechanical ventilation, APACHE II) are related to a higher mortality risk. The SSC bundles (resuscitation bundles) are associated to a protective effect (i.e. antibiotics administration, performance of haemocultures, administration of volume and vasoactive drugs and so on).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 92. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataComparison with other shrinkage methods The predictive ability of the RVM has also been compared to that of other well established shrinkage methods for regression.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 93. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataComparison with other shrinkage methods The predictive ability of the RVM has also been compared to that of other well established shrinkage methods for regression. The selected attributes and coefficients for each method were: Lasso: Age (w1 = 0.007) Ridge Regression: Germ Class (w2 = 0.005) Logistic Regression with backward feature selection: Number of dysfunctional PaO2 /FiO2 (w3 = 0.001) organs for SOFA 3-4 APACHE II (w4 = −0.006) Intercept (w1 = 4.20) (w1 = −0.021) Number of Dysfunctional SvcO2 6h (w5 = −0.001) APACHE II (w2 = −0.127) Organs (w1 = −0.12) Haematocrit 6h Worst Lactate (w6 = 0.009) APACHE II (w2 = −0.08) (w3 = −0.126). Worst Lactate (w3 = −0.25) Worst Lactate (w7 = −0.023) SvcO2 (w8 = −0.006).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 94. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed Data The three shrinkage methods evaluated in this section agreed in detecting as prognostic factors the Severity measured by the APACHE II score and acidosis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 95. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed Data The three shrinkage methods evaluated in this section agreed in detecting as prognostic factors the Severity measured by the APACHE II score and acidosis. Organ dysfunction and mechanical ventilation or other parameters related to it like PaO2 /FiO2 also play a role in the prognosis of Sepsis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 96. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed Data The three shrinkage methods evaluated in this section agreed in detecting as prognostic factors the Severity measured by the APACHE II score and acidosis. Organ dysfunction and mechanical ventilation or other parameters related to it like PaO2 /FiO2 also play a role in the prognosis of Sepsis. The accuracy of each method was the following: Method AUC Error Rate Sens. Spec. RVM 0.86 0.18 0.67 0.87 Logistic 0.75 0.30 0.64 0.72 Ridge 0.70 0.25 0.63 0.79 Lasso 0.70 0.32 0.67 0.68Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 97. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with Generative Kernels - QBK - x1 is the Number of Dysfunctional Organs as measured by the SOFA Score. x2 corresponds to Mechanical Ventilation (yes/no). x3 corresponds to Severity as Measured by the APACHE II Score. x4 corresponds to the SSC Resuscitation Bundles (i.e. administration of antibiotics, performance of haemocultures and so on). This is also a binary variable.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 98. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with Generative Kernels - QBK -     1, x4 , x3 , x3 x4 ,        x2 , x2 x4 , x2 x3 , x2 x3 x4 ,       2 2 2 2 x2 , x2 x4 , x2 x3 , x2 x3 x4 ,      3 3 3 3      x2 , x2 x4 , x2 x3 , x2 x3 x4 ,    4, x 4x , x 4x , x 4x x ,   x2 2 4 2 3 2 3 4       ESTτ = 5 5 6 x2 , x2 x4 , x2 , x1 , x1 x4 , x1 x3 , x1 x3 x4 , x1 x2 ,            2 x1 x2 x4 , x1 x2 x3 , x1 x2 x3 x4 , x1 x2 ,         2 2 2 3 x1 x2 x4 , x1 x2 x3 , x1 x2 x3 x4 , x1 x2 ,         3x , x x 3x , x x 3x x , x x 4, x1 x2 4 1 2 3 1 2 3 4 1 2      4 4 5    x1 x2 x4 , x1 x2 x3 , x1 x2 Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 99. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with Generative Kernels We have used Matlab’s Support Vector Machine QP solver implemented in the BioInformatics and Optimization Toolboxes. We have also used 10-fold cross validation to evaluate the classification performance for the different kernels. A grid search yielded the appropriate values for C parameters for each Kernel. More particularly, Quotient Basis and Fisher C = 1. Generative Kernels C = 10. Also the parameter t for the Exponential and Inverse Kernels was set to 2. Gaussian, Linear and Polynomial Kernels C = 10.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 100. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsRisk of Death Assessment from Observed DataRoD with Generative Kernels Kernel AUC Error Rate Sens. Spec. CPU time [s] Quotient 0.89 0.18 0.70 0.86 1.45 Fisher 0.76 0.18 0.68 0.86 1.39 Exponential 0.75 0.21 0.70 0.82 1.64 Inverse 0.62 0.22 0.70 0.82 1.68 Centred 0.75 0.21 0.70 0.82 1.99 Gaussian 0.83 0.24 0.65 0.81 1.56 Poly (order 2) 0.69 0.28 0.71 0.76 1.59 Linear 0.62 0.26 0.62 0.78 1.35Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 101. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContents 1 Introduction Introduction Thesis Objectives 2 Database Description Dataset Available Data 3 State of the Art 4 AI Methods Applied 5 An AI Tour of Sepsis Incidence of Sepsis Protection against Sepsis Mortality Prediction with a Latent Data Representation Risk of Death Assessment from Observed Data 6 Conclusions Incidence of Sepsis and Coadjuvant Factors Protective Effects of Statins Mortality Predictors and Their Accuracy Contributions Outline for Future Work PublicationsVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 102. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of Sepsis and Coadjuvant FactorsIncidence of Sepsis and Coadjutant Factors SIRS pathology has proven to be a very sensitive indicator of Sepsis but also one of poor specificity.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 103. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of Sepsis and Coadjuvant FactorsIncidence of Sepsis and Coadjutant Factors SIRS pathology has proven to be a very sensitive indicator of Sepsis but also one of poor specificity. Castilla y Le´n report an incidence of 250 cases /100.000 hab. o and Madrid 141 cases/100.000 hab.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 104. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of Sepsis and Coadjuvant FactorsIncidence of Sepsis and Coadjutant Factors SIRS pathology has proven to be a very sensitive indicator of Sepsis but also one of poor specificity. Castilla y Le´n report an incidence of 250 cases /100.000 hab. o and Madrid 141 cases/100.000 hab. Our Bayes Network yielded an estimation of 164 cases / 100.000 hab (i.e. 164 vs 118 in our database).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 105. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of Sepsis and Coadjuvant FactorsIncidence of Sepsis and Coadjutant Factors SIRS pathology has proven to be a very sensitive indicator of Sepsis but also one of poor specificity. Castilla y Le´n report an incidence of 250 cases /100.000 hab. o and Madrid 141 cases/100.000 hab. Our Bayes Network yielded an estimation of 164 cases / 100.000 hab (i.e. 164 vs 118 in our database). There are different comorbidities and coadjuvant factors that clearly play a role in Sepsis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 106. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of Sepsis and Coadjuvant FactorsIncidence of Sepsis and Coadjutant Factors SIRS pathology has proven to be a very sensitive indicator of Sepsis but also one of poor specificity. Castilla y Le´n report an incidence of 250 cases /100.000 hab. o and Madrid 141 cases/100.000 hab. Our Bayes Network yielded an estimation of 164 cases / 100.000 hab (i.e. 164 vs 118 in our database). There are different comorbidities and coadjuvant factors that clearly play a role in Sepsis. Two important factors to be taken into consideration are surgery or the infectious diseases such as pneumonia.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 107. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsIncidence of Sepsis and Coadjuvant FactorsIncidence of Sepsis and Coadjutant Factors SIRS pathology has proven to be a very sensitive indicator of Sepsis but also one of poor specificity. Castilla y Le´n report an incidence of 250 cases /100.000 hab. o and Madrid 141 cases/100.000 hab. Our Bayes Network yielded an estimation of 164 cases / 100.000 hab (i.e. 164 vs 118 in our database). There are different comorbidities and coadjuvant factors that clearly play a role in Sepsis. Two important factors to be taken into consideration are surgery or the infectious diseases such as pneumonia. These two factors play a very important role in the dataset analysed.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 108. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtective Effects of StatinsProtective Effects of Statins We have studied the role of pre-admission use of Statins in the incidence of Septic Shock and Prognosis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 109. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtective Effects of StatinsProtective Effects of Statins We have studied the role of pre-admission use of Statins in the incidence of Septic Shock and Prognosis. This has been done through Graphical Models (Algebraic Statistics) and Regression Trees.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 110. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtective Effects of StatinsProtective Effects of Statins We have studied the role of pre-admission use of Statins in the incidence of Septic Shock and Prognosis. This has been done through Graphical Models (Algebraic Statistics) and Regression Trees. There is a clear dependence between pre-admission use of statins the outcome of Sepsis.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 111. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtective Effects of StatinsProtective Effects of Statins We have studied the role of pre-admission use of Statins in the incidence of Septic Shock and Prognosis. This has been done through Graphical Models (Algebraic Statistics) and Regression Trees. There is a clear dependence between pre-admission use of statins the outcome of Sepsis. We have seen that the higher the severity and organ dysfunction, the higher the protection.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 112. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsProtective Effects of StatinsProtective Effects of Statins We have studied the role of pre-admission use of Statins in the incidence of Septic Shock and Prognosis. This has been done through Graphical Models (Algebraic Statistics) and Regression Trees. There is a clear dependence between pre-admission use of statins the outcome of Sepsis. We have seen that the higher the severity and organ dysfunction, the higher the protection. There is no clear dependence between statins and the incidence of Septic Shock.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 113. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Predictors and Their AccuracyMortality Predictors and Their Accuracy The main limitation of current indicators for scoring the evolution of sepsis is their lack of specificity.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 114. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Predictors and Their AccuracyMortality Predictors and Their Accuracy The main limitation of current indicators for scoring the evolution of sepsis is their lack of specificity. Not only does this affect incidence rates but also prognosis since many patients are given treatment that is not required.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 115. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Predictors and Their AccuracyMortality Predictors and Their Accuracy The main limitation of current indicators for scoring the evolution of sepsis is their lack of specificity. Not only does this affect incidence rates but also prognosis since many patients are given treatment that is not required. We have analysed 17 different approaches to assess RoD and compared them with standard practice (APACHE II).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 116. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Predictors and Their AccuracyMortality Predictors and Their Accuracy The main limitation of current indicators for scoring the evolution of sepsis is their lack of specificity. Not only does this affect incidence rates but also prognosis since many patients are given treatment that is not required. We have analysed 17 different approaches to assess RoD and compared them with standard practice (APACHE II). The set of variables selected through RVM are consistent with clinical practice and the SSC guidelines.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 117. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsMortality Predictors and Their AccuracyMortality Predictors and Their Accuracy Method AUC Error Rate Sens. Spec. Dataset LR-FA 0.78 0.24 0.65 0.80 FA LR 0.75 0.30 0.64 0.72 LR APACHE II 0.70 0.28 0.82 0.55 N/A RVM 0.86 0.18 0.67 0.87 RVM Ridge 0.70 0.25 0.63 0.79 RVM Lasso 0.70 0.32 0.67 0.68 RVM SVM-Quotient 0.89 0.18 0.70 0.86 RVM SVM-Fisher 0.76 0.18 0.68 0.86 RVM SVM-EXP 0.75 0.21 0.70 0.82 RVM SVM-INV 0.62 0.22 0.70 0.82 RVM SVM-CENT 0.75 0.21 0.70 0.82 RVM SVM-GAUSS 0.83 0.24 0.65 0.81 RVM SVM-LIN 0.62 0.26 0.62 0.78 RVM SVM-POLY 0.69 0.28 0.71 0.76 RVMVicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 118. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContributionsContributions The application of Algebraic Models and the study of Quotient Basis resulted in the definition of the Quotient Basis Kernel.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 119. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContributionsContributions The application of Algebraic Models and the study of Quotient Basis resulted in the definition of the Quotient Basis Kernel. The application of Algebraic Models also resulted in the definition of a simplified version of the Fisher kernel.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 120. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContributionsContributions The application of Algebraic Models and the study of Quotient Basis resulted in the definition of the Quotient Basis Kernel. The application of Algebraic Models also resulted in the definition of a simplified version of the Fisher kernel. We have provided a set of actionable ROD indicators for Severe Sepsis, which are readily interpretable and used in an ICU setting (LR and LR-FA).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 121. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsContributionsContributions The application of Algebraic Models and the study of Quotient Basis resulted in the definition of the Quotient Basis Kernel. The application of Algebraic Models also resulted in the definition of a simplified version of the Fisher kernel. We have provided a set of actionable ROD indicators for Severe Sepsis, which are readily interpretable and used in an ICU setting (LR and LR-FA). Preadmission use of Statins for septic patients is closely related to severity and organ dysfunction.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 122. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsOutline for Future WorkOutline for Future Work Generalise the methods and algorithms proposed in this Ph.D. to other general datasets.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 123. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsOutline for Future WorkOutline for Future Work Generalise the methods and algorithms proposed in this Ph.D. to other general datasets. Study Sepsis from a proteomics point of view (i.e. identification of biomarkers for Sepsis or other inflammatory mediators).Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 124. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsOutline for Future WorkOutline for Future Work Generalise the methods and algorithms proposed in this Ph.D. to other general datasets. Study Sepsis from a proteomics point of view (i.e. identification of biomarkers for Sepsis or other inflammatory mediators). Exploit the algebraic relation between the FA model and DBN-RBM to obtain efficient algorithms to train the latter.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 125. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsOutline for Future WorkOutline for Future Work Generalise the methods and algorithms proposed in this Ph.D. to other general datasets. Study Sepsis from a proteomics point of view (i.e. identification of biomarkers for Sepsis or other inflammatory mediators). Exploit the algebraic relation between the FA model and DBN-RBM to obtain efficient algorithms to train the latter. Study the algebraic relation between the simplified Fisher kernel and the QBK.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 126. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsPublicationsPublications Ribas, V., Ruiz-Rodr´ıguez, J.D., Wojdel, A., Caballero-L´pez, J., Ruiz-Sanmart´ A., Rello, J. and Vellido, o ın A. Severe sepsis mortality prediction with Relevance Vector Machines. In Procs. of the 33rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC 2011). Ribas, V.J., Caballero-L´pez, J., Saez de Tejada, A., Ruiz-Rodr´ o ˜ ıguez, J.C., Ruiz-SanmartAn, A., Rello, J., Vellido, A. Graphical models for ICU outcome prediction in sepsis patients treated with statin drugs, In Procs. of the Eigth International Meeting on Computational Intelligence Methods in Bioinformatics and Biostatistics, (CIBB 2011). Ribas, V., Caballero-L´pez, J., Ruiz-Rodr´ o ıguez, J.C., Ruiz Sanmart´ A., Rello, J., and Vellido, A. On the ın, use of decision trees for ICU outcome prediction in sepsis patients treated with statins. In Procs. of the IEEE Symposium Series on Computational Intelligence / IEEE Symposium on Computational Intelligence and Data Mining (IEEE SSCI CIDM 2011), pp.37-43. Ribas, V.J, Vellido, A., Ruiz-Rodr´ıguez, J.C., Intelligent Management of Sepsis in the Intensive Care Unit in Intelligent Data Analysis for Real-Life Applications: Theory and Practice, IGI pub., in press. Ribas V.J., Romero E., Ruiz-Rodr´ ıguez, J.C., Vellido A., A Quotient Basis Kernel for the prediction of mortality in severe sepsis patients, ESANN 2013. Accepted. Ribas V.J., Romero E., Ruiz-Rodr´ ıguez, J.C., Vellido A., A Quotient Basis Kernel for the prediction of mortality in severe sepsis patients, ESANN 2013. Accepted. Ribas V.J., Vellido A., Romero E., Ruiz-Rodr´ ıguez, J.C., Sepsis Mortality Prediction with Quotient Basis, Medical & Biological Eng & Computing (MBEC), Springer. Submitted.Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit
  • 127. Introduction Database Description State of the Art AI Methods Applied An AI Tour of Sepsis ConclusionsPublications THANK YOU!!Vicent J. Ribas LSI - SOCO Technical University of Catalonia (UPC) BarcelonaOn the Intelligent Management of Sepsis in the Intensive Care Unit