Computational Physiology for Critical Care Monitoring

1
Computational Physiology for Critical Care
Monitoring
Stuart Russell, UC Berkeley Joint work with Geoff
Manley, Mitch Cohen, Kristan Staudenmayer, Diane
Morabito (UCSF), Norm Aleks, Nimar Arora, Shaunak
Chatterjee (UCB)
2
(No Transcript)
3
Critical care

• 300B/yr in US, high morbidity/mortality
• Goal improve outcomes, reduce length of stay, do
  science
• Approach
• Large-scale data repository for worldwide
  research use
• Currently 60GB, 16 ICU beds monitored 24/7, soon
  multi-institutional
• First release any day now .
• Data mining for outcome prediction, early
  warning, etc.
• Real-time model-based estimation of patient state
• (And systems physiology model-building)

4
Critical care state estimation

• Given
• 140 initial presentation fields
• 40 real-time sensor streams
• 1500 asynchronous measures (blood, drugs, etc.)
• Compute posterior probability distribution for
• 100 (patho)physiological state variables
• Method
• Patient-adaptive dynamic Bayesian network (DBN)
  stochastic models of physiology and sensor
  dynamics (c.f. Guyton et al., 1972, 354-variable
  nonlinear ODE)
• Flexible across time scales, models, sensors
  (images, text, etc.)
• Can incorporate genetic factors (observed or
  unobserved)

5
Human physiology v0.1
6
Medullary cardiovascular center
Cardiac parasympathetic output
Cardiac sympathetic output
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Heart rate
Cardiac contrac-tility
Venous tone
Arterio-lar tone
Cardiac preload
Capillary pressure
Cardiac stroke volume
Cardiac output
Vascular resistance
Mean arterial blood pressure
Barorecep-tor discharge
7
Medullary cardiovascular center
Medullary cardiovascular center
Cardiac parasympathetic output
Cardiac sympathetic output
Cardiac parasympathetic output
Cardiac sympathetic output
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Heart rate
Cardiac contrac-tility
Venous tone
Arterio-lar tone
Heart rate
Cardiac contrac-tility
Venous tone
Arterio-lar tone
Cardiac preload
Capillary pressure
Cardiac preload
Capillary pressure
Cardiac stroke volume
Cardiac stroke volume
Cardiac output
Vascular resistance
Cardiac output
Vascular resistance
Mean arterial blood pressure
Mean arterial blood pressure
Barorecep-tor discharge
Barorecep-tor discharge
8
Setpoint inputs from ANS, CNS, intracranial,
blood
Setpoint inputs from ANS, CNS, intracranial,
blood
Medullary cardiovascular center
Medullary cardiovascular center
Cardiac parasympathetic output
Cardiac sympathetic output
Cardiac parasympathetic output
Cardiac sympathetic output
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Heart rate
Cardiac contrac-tility
Venous tone
Blood volume
Arterio-lar tone
Pulm. intra-thoracic press.
Heart rate
Cardiac contrac-tility
Venous tone
Blood volume
Arterio-lar tone
Pulm. intra-thoracic press.
Cardiac preload
Capillary pressure
Cardiac preload
Capillary pressure
Cardiac stroke volume
Cardiac stroke volume
Blood transu-dation
Blood transu-dation
Cardiac output
Vascular resistance
Cardiac output
Vascular resistance
Mean arterial blood pressure
Mean arterial blood pressure
Barorecep-tor discharge
Intracranial physiology
Tissues-NOS perfusion
GI/Liver perfusion
Barorecep-tor discharge
Intracranial physiology
Tissues-NOS perfusion
GI/Liver perfusion
9
Setpoint inputs from ANS, CNS, intracranial,
blood
Setpoint inputs from ANS, CNS, intracranial,
blood
Medullary cardiovascular center
Medullary cardiovascular center
Cardiac parasympathetic output
Cardiac sympathetic output
Cardiac parasympathetic output
Cardiac sympathetic output
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Heart rate
Cardiac contrac-tility
Venous tone
Blood volume
Arterio-lar tone
Pulm. intra-thoracic press.
Heart rate
Cardiac contrac-tility
Venous tone
Blood volume
Arterio-lar tone
Pulm. intra-thoracic press.
Cardiac preload
Capillary pressure
Cardiac preload
Capillary pressure
Heart rate sensor model
Heart rate sensor model
Cardiac stroke volume
Cardiac stroke volume
Central venous pressure sensor model
Central venous pressure sensor model
Blood transu-dation
Blood transu-dation
Cardiac output
Vascular resistance
Cardiac output
Vascular resistance
Mean arterial blood pressure
Mean arterial blood pressure
MAP sensor model
MAP sensor model
Barorecep-tor discharge
Intracranial physiology
Tissues-NOS perfusion
GI/Liver perfusion
Barorecep-tor discharge
Intracranial physiology
Tissues-NOS perfusion
GI/Liver perfusion
10
Setpoint inputs from ANS, CNS, intracranial,
blood
Setpoint inputs from ANS, CNS, intracranial,
blood
Medullary cardiovascular center
Medullary cardiovascular center
PK conc. of phenyl-ephrine
PK conc. of phenyl-ephrine
Cardiac parasympathetic output
Cardiac sympathetic output
Cardiac parasympathetic output
Cardiac sympathetic output
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Card. M2
Card. ß1
Card. ß2
Vasc. a1
Vasc. a2
Vasc. ß2
Heart rate
Cardiac contrac-tility
Venous tone
Blood volume
Arterio-lar tone
Pulm. intra-thoracic press.
Heart rate
Cardiac contrac-tility
Venous tone
Blood volume
Arterio-lar tone
Pulm. intra-thoracic press.
Cardiac preload
Capillary pressure
Cardiac preload
Capillary pressure
Heart rate sensor model
Heart rate sensor model
Cardiac stroke volume
Cardiac stroke volume
Central venous pressure sensor model
Central venous pressure sensor model
Blood transu-dation
Blood transu-dation
Cardiac output
Vascular resistance
Cardiac output
Vascular resistance
Mean arterial blood pressure
Mean arterial blood pressure
MAP sensor model
MAP sensor model
Barorecep-tor discharge
Intracranial physiology
Tissues-NOS perfusion
GI/Liver perfusion
Barorecep-tor discharge
Intracranial physiology
Tissues-NOS perfusion
GI/Liver perfusion
11
Real data are messy
12
(No Transcript)
13
(No Transcript)
14
ALARM
15
(No Transcript)
16
Next Steps

• Reduce ICU false alarms from gt90 to lt5
• Demonstrate clinically relevant inferences,
  e.g.,
• Vascular stiffness
• Erroneous drug administration
• Pulmonary artery pressure (w/o catheter!)
• Extend physiology model to all major systems
• Multiscale connect physiology to molecules