Prognosis Modeling

• Process of Diagnosis
  • Integration of Data and Expert Observation & Knowledge
    • Health record (medical history, common "chart" data)
    • Observation (by trained experts)
    • Results from specialized diagnostic tests (e.g., images)
    • Integration with Evidence-Based Knowledge
  • Levels of Diagnosis
    • Classification from medical/disease model framework
    • Classification via ICF model framework
• Process of Determining Prognosis (through expert inference)
  • Definitions for Prognosis
    • Typically of interest: prediction of steady-state status
    • Prognosis is function of intervention plan and its implementation
  • Dynamics of Healing/Recovery Processes
    • Spontaneus recovery mechanisms
      • tissue matrix, intracellular mechanisms, "macrosystems" role
    • Healing/remodeling of connective tissue (ligaments, skin, ...)
    • Healing/remodeling of hard tissue (bone) and cartilage
    • Healing of skeletal muscle tissue
    • Healing of "neural tissue"
  • Prognosis made by Trained Expert(s)
  • Effects of Available Resources (for allocation)
    • Relative to spontaneous recovery
    • Maintenance of protocols that assist spontaneous mechanisms
    • Anticipated imapct of various intervention plans
  • Effects of Ongoing Assessments
    • Prognosis may change as new data becomes available
    • Adjustment in prognosis may affect intervention plan
      • if so, closed-loop system (sampled-data)
• Systems Model of Expert Prognosis (Prediction)
  • Inputs
    • Status data (e.g., from records, observation)
      • general patient data (both history, present)
      • results from diagnostic tests (past, present)
      • present diagnosis
        • sets classification used for variables, rules
    • Events
      • Interventions
      • Other health-related events
  • States
    • Variables representing the "state" of person
    • Change with time (with roughly known time scale)
    • Function of inputs and states
  • Outputs
    • Measures of performance, capabilities
    • Function of inputs and states
    • Often experimentally measurable
  • Outcomes:
    • Measures to be maximized/minimized
      • in optimization: "performance criterion" or "cost function"
    • Function of inputs, states, outputs
    • Key "global" outcomes may be function of other sub-outcomes
    • Ideally experimentally measurable
    • Key focus on prognosis process

Intelligent Telerehab Assistant for Prognosis (ITA-Predict)

• Part of Doctoral research of Yu Wang ("Fish")
• Mathematical Model for Dynamic Prognosis Prediction
• Uses "systems" modeling approach - inputs, states, outputs
• Uses fuzzy inference to extract expert reasoning
  • membership functions to map to "logic/inference" world of expert
• Fuzzy rules used to capture causality of how states can change
  • if [ , , ,] then [state ...]
  • experts create rules
• Use of simulation and sensitivity analysis tools to help experts refine membership functions, rules