- Email: [email protected]
A computerized decision support system to predict the variations in the cerebral blood flow of mechanically ventilated infants
Computers in Biology and Medicine 43 (2013) 1402–1406
Contents lists available at ScienceDirect
Computers in Biology and Medicine journal homepage: www.elsevier.com/locate/cbm
A computerized decision support system to predict the variations in the cerebral blood flow of mechanically ventilated infants Fleur T. Tehrani n Department of Electrical Engineering, California State University, 800 North State College Boulevard, Fullerton, CA 92831, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 21 May 2013 Accepted 26 June 2013
A computerized decision support system is described to predict the changes in the cerebral blood flow (CBF) of mechanically ventilated infants in response to different ventilatory settings. A CBF controller was developed and combined with a mathematical model of the infant's respiratory system to simulate the effects of ventilatory settings on the infant's CBF. The performance of the system was examined under various ventilatory treatments and the results were compared with available experimental data. The comparisons showed good agreement between the simulation results and experimental data for preterm infants. These included the results obtained under conditions of hypoventilation, hyperventilation, hypoxia, and hyperoxia. The presented decision support system has the potential to be used as an aide to the intensivist in choosing appropriate ventilation treatments for infants to prevent the untoward consequences of hazardous changes in CBF in mechanically ventilated infants such as hypoxic-ischemic brain injuries. & 2013 Elsevier Ltd. All rights reserved.
Keywords: Decision support systems Computer modeling Cerebral blood flow Mechanical ventilation Infant
1. Introduction Newborn premature infants often suffer from various respiratory disorders [1–4] and may need mechanical ventilation assistance during the first few weeks of life. However, mechanical ventilation treatment can have many side effects and may cause serious complications in neonates which include development of pulmonary interstitial emphysema, broncho-pulmonary dysplasia (BPD), and retinopathy of prematurity [5,6]. Cerebrovascular injuries are among the most serious complications that can develop during the first week of life of these infants and such complications can be caused by cerebral hemorrhage and ischemic injuries to the brain [7,8]. Disturbances in the cerebral blood flow (CBF) beyond the normal range can cause cerebrovascular injuries that may in turn lead to long-term disabilities or death among premature neonates. CBF is the ratio of the cerebral perfusion pressure to cerebrovascular resistance. Therefore, physical factors that can change these two parameters can affect CBF. Moderate changes in the mean arterial blood pressure (MABP) have not been found to cause significant changes in CBF of infants with intact cerebral autoregulation [9]. If abrupt changes in the MABP are prevented, and in the absence of hypoglycemia and anemia that can increase CBF, the most important factors affecting the CBF of mechanically ventilated infants with intact CBF autoregulation are considered
n
Tel.: +1 714 281 5859; fax: +1 714 281 1360. E-mail address: [email protected]
0010-4825/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.compbiomed.2013.06.015
to be the arterial partial pressures of carbon dioxide and oxygen [7–9]. Therefore, proper regulation of infant's blood gases through appropriate selection of ventilator settings can significantly help to reduce hazardous disturbances of CBF that can lead to periventricular hemorrhage and hypoxic-ischemic brain injuries in premature infants. In this article, a decision support system (DSS) is presented by which the effects of different ventilator treatments on infants' CBF are simulated by using a mathematical model. At this time, continuous monitoring systems utilizing non-invasive technologies such as near-infrared spectroscopy and Doppler ultrasound are available that can be used to measure or estimate the changes in the infants' CBF [8,10]. However, routine clinical application of these systems is questionable due to their limitations and practical problems [8]. In the absence of direct CBF monitoring systems, or in conjunction with such support technologies, the technique described in this article can be used at bedside to help the intensivist choose appropriate ventilator settings for infants on mechanical ventilation and avoid hazardous disturbances in CBF and their untoward consequences. 2. Methodology A block diagram of the mathematical system used in the DSS is shown in Fig. 1. In this system, chemical control of infant's CBF is combined with a mathematical model of the infant's respiratory system [11,12] to simulate the effects of different ventilation treatments on the infant's CBF. The simulation results are
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
1403
Fig. 2. %ΔCBF as a function of P aCO2 .
Fig. 3. %ΔCBF as a function of P aO2 . Fig. 1. A block diagram of the computerized system to predict changes in the CBF of infants in response to ventilatory settings.
compared to reported clinical studies to assess the effectiveness of the system.
due to changes in P aCO2 . Combining Eqs. (2) and (3) yields
2.1. The CBF controller To design the CBF controller, experimental data from the literature were compiled to derive mathematical relations between the infants' CBF and arterial blood gases. In compiling and analyzing the experimental data, care was taken to choose the results obtained when both the arterial partial pressure of carbon dioxide, P aCO2 , and the arterial partial pressure of oxygen, P aO2 , were monitored at the same time. For the CBF response to P aCO2 , the results reported in several references were analyzed [13–16]. To measure the changes in CBF in response to P aCO2 , Doppler ultrasound techniques were used in References [13–15], and the intravenous Xe clearance technique was employed in the experiments of Reference [16]. The experimental data obtained for mechanically ventilated infants with gestational ages ranging from 26 to 33 weeks were selected. The median gestational age in this group was 29 weeks. For the CBF response to changes in P aO2 , the experimental data reported in References [15,17] were analyzed. In both studies, Doppler ultrasound techniques were used to measure CBF. Experimental data obtained for infants under 37 weeks of gestation were used for analysis. The median gestational age of these infants was 31 weeks ranging from 25 to 36 weeks. Percentage variations in CBF in comparison to normal CBF values were obtained and plotted as functions of P aCO2 and P aO2 . Normal CBF was considered at P aCO2 ¼40 mm Hg, and P aO2 ¼100 mm Hg. Figs. 2 and 3 show the percentage variations in infants' CBF (%ΔCBF) as compared to normal versus P aCO2 , and P aO2 respectively. Curve fitting techniques and linear regression analyses were performed on the data presented in Figs. 2 and 3. The results of the linear regression analyses are %ΔCBF ¼ 3:67ðP aCO2 –40Þ þ 2:257
due to changes in P aO2 (with r ¼0.90698).
2 %ΔCBF ¼ 0:0933½ΔP aO2 þ 3:69354½ΔP aCO2 0:00458 ΔP aCO2 þ 3:29354
ð4Þ
due to changes in blood gases where ΔP aO2 ¼ 100 P aO2 and ΔP aCO2 ¼P aCO2 40, and P aO2 and P aCO2 are in mm Hg. Eq. (4) gives the steady state value of %ΔCBF in relation to changes in arterial blood gases. In the controller, the steady state value of CBF is found by adding %ΔCBF from Eq. (4) to normal CBF, and the instantaneous value of CBF is determined by adding a 1st order lag with a time constant of 5 s. 2.2. The mathematical model of the infant's respiratory system The other building blocks of the system in Fig. 1 consisting of Lungs, Body Tissue, Transport Delay, and the Brain Tissue are parts of a mathematical model of the infant's respiratory system [12]. This detailed model of the human respiratory system has been used by a number of researchers to simulate the neonatal respiratory system e.g., [18–20]. The peripheral and central respiratory receptors and the respiratory controller of the model were not included in the system of Fig. 1 because in this DSS the respiratory drive signal is not provided by the infant's own respiratory control centers and is coming from the ventilator. The effects of an anatomical shunt in the lung is included in the system and the lung volume is time varying. The mathematical equations describing different building blocks of the model that have been incorporated in the DSS of Fig. 1 have been presented in Reference [12] and are not repeated here for brevity.
ð1Þ
3. Results
ð2Þ
Figs. 4–9 show a series of simulation results by using the presented DSS. In all the simulation tests, the patient is modeled as a premature infant of 1.8 kg weight, with respiratory distress syndrome (RDS), and a respiratory dynamic compliance of 2 ml/
due to changes in P aCO2 (with r ¼0.85398) and %ΔCBF ¼ 0:0933ð100P aO2 Þ þ 0:876
With regard to %ΔCBF response to P aCO2 , a quadratic fit was found to be more accurate than a linear formula. That quadratic equation was found as 2 %ΔCBF ¼ 2:41754 þ 3:69354ðP aCO2 40Þ0:00458 P aCO2 –40 ð3Þ
1404
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
Fig. 4. A set of simulation results of the DSS. The infant's ventilation and oxygenation are normal.
Fig. 6. A set of simulation results of the DSS in response to mild hyperventilation. The infant's oxygenation is normal.
Fig. 5. A set of simulation results of the DSS. The infant's oxygenation is normal while ventilation is low, resulting in hypercapnia.
Fig. 7. A set of simulation results of the DSS in response to low oxygen breathing.
cm H2O. A lung shunt ratio of 10% is assumed and the level of positive end-expiratory pressure (PEEP) is 5 cm H2O in all the tests.
Fig. 4a and b shows the simulation results with a respiratory frequency set at 40 breaths/min, minute ventilation at 0.59 l/min, and an inspired fraction of oxygen, F IO2 , set at 22.5%. Fig. 4a shows
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
1405
In Fig. 6, F IO2 is set at 22% and minute ventilation is raised to 0.68 l/min to induce hypocapnia. Fig. 7 shows the simulation results of the DSS with minute ventilation at 0.59 l/min, and F IO2 set at 18.5% to assess the effects of induced hypoxemia. Fig. 8 shows the simulation results of blood gases and CBF in response to induced hyperoxemia with minute ventilation at 0.59 l/min and F IO2 set at 28%. Fig. 9 shows the simulation results of the DSS with combined effects of hypoxemia and hypercapnia. In this test, minute ventilation is set at 0.5 l/min, and F IO2 is at the ambient air concentration of 21%.
4. Discussion
Fig. 8. A set of simulation results of the DSS in response to hyperoxia.
Fig. 9. A set of simulation results of the DSS in response to combined effects of hypoventilation and mild hypoxia.
the simulation results of the infant's P aO2 , and P aCO2 , while Fig. 4b shows the CBF predictions of the DSS. Fig. 5a and b shows the simulation results of blood gases and CBF for the same infant with F IO2 at 25%, and minute ventilation reduced to 0.5 l/min to induce hypercapnia.
The results of Fig. 4a shows that with the selected ventilatory settings, P aO2 rises close to 98 mm Hg and P aCO2 remains around 40 mm Hg over time. Fig. 4b shows the infant's CBF response which rises by about 9.5% initially due to a slight transient increase in P aCO2 and the initial low value of P aO2 . As shown in Fig. 4b, after an initial rise, CBF falls with time and approaches its normal level of 3.33 ml/s (corresponding to 200 ml/min or 11.1 ml/100 g/min) as blood gases approach and remain within normal limits. The results of Fig. 5a show an increase in P aCO2 over time to close to 50 mm Hg due to hypoventilation. The P aO2 value remains around 103 mm Hg during most of the time due to oxygen supplementation. Fig. 5b shows the CBF response due to induced hypercapnia. It is seen that CBF rises steadily to about 36% higher than its initial value to about 4.54 ml/s, corresponding to around 15 ml/100 g/min in 100 min of simulation in response to hypercapnia. The P aO2 level remains normal and does not cause any change in CBF in this test. These results are in good agreement with experimental findings [9,14,16]. Fig. 6a shows the simulation blood gas results of the DSS when the infant's ventilation is increased to 0.68 l/min to induce hypocapnia. The F IO2 level is kept near the ambient oxygen concentration at 22% to avoid hyperoxemia. As seen in Fig. 6a, P aO2 remains normal at about 100 mm Hg and P aCO2 falls to around 32.5 mm Hg over time as a result of hyperventilation. Fig. 6b shows the infant's CBF response to the induced hypocapnia. As seen, CBF falls to around 2.53 ml/s (151.8 ml/min corresponding to 8.43 ml/100 g/ min) with time. This 24% reduction in CBF in response to about 7.5 mm Hg drop in P aCO2 is in good agreement with experimental observations [16]. Fig. 7a shows the DSS's simulation results of the infant's blood gases with imposed low oxygen breathing. As seen, P aCO2 remains close to 40 mm Hg due to adequate ventilation settings but P aO2 falls to about 68.5 mm Hg with time due to low oxygen breathing. Fig. 7b shows the CBF response of the DSS in this test. As seen, there is an initial 11% rise in CBF to about 3.7 ml/s (222 ml/min or 12.3 ml/100 g/min) due to the initial rapid fall of P aO2 to about 65 mm Hg while P aCO2 rises to around 41.5 mm Hg. As P aO2 stabilizes around 68.5 mm Hg and P aCO2 decreases by a few mm Hg during the test, CBF approaches 3.45 ml/s that corresponds to 11.5 ml/100 g/min for this infant over time. This represents a final increase of about 3.6% in CBF due to the induced hypoxemia. These results show general agreement with the experimental data [17]. Fig. 8a shows the P aO2 and P aCO2 results of the DSS when oxygen supplementation provided to the infant is increased (F IO2 is 28%). This figure shows that P aCO2 remains around a normal value of 40 mm Hg while P aO2 rises to about 135 mm Hg over time. Fig. 8b shows that the infant's CBF falls to about 3.22 ml/s (i.e., 10.7 ml/ 100 g/min) in this test. This represents a fall of 3.3% in CBF due to the induced hyperoxemia. There is general agreement between these results and the experimental observations in preterm infants
1406
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
[17]. For normal infants however, the sensitivity of the CBF response to changes in P aO2 may be significantly higher than in premature infants as reported in the literature [17]. In Fig. 9a and b, the simulation results of the DSS are shown when ventilation is reduced to 0.5 l/min and no oxygen supplementation is provided to the infant (F IO2 is 21%). As a result of these ventilation settings, P aCO2 rises to about 48.5 mm Hg due to hypoventilation and P aO2 stabilizes around 75 mm Hg over time. Fig. 9b shows a steady increase in the infant's CBF response to about 4.6 ml/s (i.e., 15.3 ml/100 g/min) in response to combined effects of hypercapnia and hypoxemia. This represents an increase of about 38% in CBF. These results are in general agreement with experimental data for premature infants [9,14,16,17]. Once again, a significantly higher sensitivity in the CBF response to variations in P aO2 may be expected for normal full-term infants [17]. 5. Summary A computerized decision support system is presented for predicting the effects of different ventilatory settings on the cerebral blood flow of premature infants. The results of this study show that the presented system can be used to estimate the effects of changes in various ventilatory settings including minute ventilation, tidal volume, respiratory rate, PEEP, and F IO2 on the cerebral blood flow of preterm infants on mechanical ventilation. In the absence of direct monitoring systems, the system presented in this article can be used to estimate the cerebral blood flow variations for infants on mechanical ventilation provided the infants have intact cerebral autoregulation. Conflict of interest statement None declared.
[6] P. Sapieha, J.S. Joyal, J.C. Rivera, E. Kermorvant-Duchemin, F. Sennlaub, P. Hardy, P. Lachapelle, S. Chemtob, Retinopathy of prematurity: understanding ischemic retinal vasculopathies at an extreme of life, J. Clin. Invest. 120 (9) (2010) 3022–3032. [7] G. Greisen, Cerebral blood flow and energy metabolism in the newborn, Clin. Perinatol. 24 (3) (1997) 531–546. [8] K.D. Liem, G. Greisen, Monitoring of cerebral haemodynamics in newborn infants, Early Hum. Dev. 86 (3) (2010) 155–158. [9] J.R. Kaiser, G.H. Gauss, D.K. Williams, Tracheal suctioning is associated with prolonged disturbances of cerebral hemodynamics in very low birth weight infants, J. Perinatol. 28 (1) (2008) 34–41. [10] A. Milan, F. Freato, V. Vanzo, L. Chiandetti, P. Zaramella, Influence of ventilation mode on neonatal cerebral blood flow and volume, Early Hum. Dev. 85 (7) (2009) 415–419. [11] F.T. Tehrani, Autoregulation of the cerebral blood flow in preterm infants, in: Proceedings of IEEE-EMBC and CMBEC Conference, 1995, pp. 635–636. [12] F.T. Tehrani, Mathematical analysis and computer simulation of the respiratory system in the newborn infant, IEEE Trans. Biomed. Eng. 40 (5) (1993) 475–481. [13] M.I. Levene, D. Shortland, N. Gibson, D.H. Evans, Carbon dioxide reactivity of the cerebral circulation in extremely premature infants: effects of postnatal age and Indomethacin, Pediatr. Res. 24 (2) (1988) 175–179. [14] A.C. Fenton, K.L. Woods, D.H. Evans, M.I. Levene, Cerebrovascular carbon dioxide reactivity and failure of autoregulation in preterm infants, Arch. Dis. Child. 67 (7_Spec_No) (1992) 835–839. [15] S.T. Kempley, H.R. Gamsu, Arterial blood pressure and blood flow velocity in major cerebral and visceral arteries—I. Interindividual differences, Early Hum. Dev. 34 (3) (1993) 227–232. [16] O. Pryds, G. Greisen, Effects of P aCO2 and haemoglobin concentration on day to day variation of CBF in preterm neonates, Acta Paediatr. Scand. 360 (1989) 33–36. [17] S. Niijima, D.B. Shortland, M.I. Levene, D.H. Evans, Transient hyperoxia and cerebral blood flow velocity in infants born prematurely and at full term, Arch. Dis. Child. 63 (10_Spec_No) (1988) 1126–1130. [18] M.A. Mananas, C. Navarro, S. Romero, R. Grino, R. Rabinovich, S. Benito, P. Caminal, Control system response of different respiratory models under ventilatory stimuli and pathologies. in: Proceeding of the 15th World Congress of the International Federation on Automatic Control (IFAC), vol. 15, 2002, pp. 2317–2322. [19] R. Farrenkopf, Automatic Control of Oxygen Concentration to Premature Neonates. VPMA Progress Reports and Papers, 2002. Available from: 〈http:// www.docstoc.com/docs/95021462/ Automatic-Control-of-Oxygen-Flow-to-Premature-Neonates〉. [20] A.M. Hernandez, M.A. Mananas, R. Costa-Castello, Learning respiratory system function in BME studies by means of a virtual laboratory: respilab, IEEE Trans. Educ. 51 (1) (2008) 24–34.
References [1] T.A. Merritt, Respiratory distress, in: M. Ziai, T.A. Clarke, T.A. Merritt (Eds.), Assessment of the Newborn, a Guide for the Practitioner, 1st ed.,Little, Brown and Company, Boston, 1984. [2] N.C. Bradburn, Neurological causes of respiratory distress, in: R.L. Schreiner, N.C. Bradburn (Eds.), Care of the Newborn, 2nd ed.,Raven Press, New York, 1988. [3] S. Farzan, Respiratory distress syndrome of the newborn: hyaline membrane disease, in: S. Farzan (Ed.), A Concise Handbook of Respiratory Distress, 2nd ed.,Reston Publishing Company Inc., A Prentice Hall Company Reston, 1985. [4] D.L. Levin, Hyaline membrane disease, in: D.L. Levin, F.C. Morriss, G.C. Moore (Eds.), A Practical Guide to Pediatric Intensive Care, The C.V. Mosby Company St. Louis, 1979. [5] F. Riedel, Long term effects of artificial ventilation in neonates, Acta Paediatr. Scand. 76 (1) (1987) 24–29.
Fleur T. Tehrani holds a B.S. degree in electrical engineering from Sharif University of Technology, Tehran, Iran, a graduate diploma from Imperial College of Science and Technology, London, and M.S. and Ph.D. degrees in electrical engineering from the University of London, UK. She has been on the faculty of California State University, Fullerton, since 1985 where she is currently a tenured professor of electrical engineering. She is the inventor of metabolic rate monitors and several respiratory assist devices including an advanced respiratory control system known as “Adaptive Support Ventilation.” Dr. Tehrani is a Fellow of the Institution of Engineering and Technology (the IET), a Fellow of the Institute for the Advancement of Engineering, a senior member of the Institute of Electrical and Electronics Engineers (IEEE), and a National Life Member of Graduate Women in Science (GWIS).
Contents lists available at ScienceDirect
Computers in Biology and Medicine journal homepage: www.elsevier.com/locate/cbm
A computerized decision support system to predict the variations in the cerebral blood flow of mechanically ventilated infants Fleur T. Tehrani n Department of Electrical Engineering, California State University, 800 North State College Boulevard, Fullerton, CA 92831, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 21 May 2013 Accepted 26 June 2013
A computerized decision support system is described to predict the changes in the cerebral blood flow (CBF) of mechanically ventilated infants in response to different ventilatory settings. A CBF controller was developed and combined with a mathematical model of the infant's respiratory system to simulate the effects of ventilatory settings on the infant's CBF. The performance of the system was examined under various ventilatory treatments and the results were compared with available experimental data. The comparisons showed good agreement between the simulation results and experimental data for preterm infants. These included the results obtained under conditions of hypoventilation, hyperventilation, hypoxia, and hyperoxia. The presented decision support system has the potential to be used as an aide to the intensivist in choosing appropriate ventilation treatments for infants to prevent the untoward consequences of hazardous changes in CBF in mechanically ventilated infants such as hypoxic-ischemic brain injuries. & 2013 Elsevier Ltd. All rights reserved.
Keywords: Decision support systems Computer modeling Cerebral blood flow Mechanical ventilation Infant
1. Introduction Newborn premature infants often suffer from various respiratory disorders [1–4] and may need mechanical ventilation assistance during the first few weeks of life. However, mechanical ventilation treatment can have many side effects and may cause serious complications in neonates which include development of pulmonary interstitial emphysema, broncho-pulmonary dysplasia (BPD), and retinopathy of prematurity [5,6]. Cerebrovascular injuries are among the most serious complications that can develop during the first week of life of these infants and such complications can be caused by cerebral hemorrhage and ischemic injuries to the brain [7,8]. Disturbances in the cerebral blood flow (CBF) beyond the normal range can cause cerebrovascular injuries that may in turn lead to long-term disabilities or death among premature neonates. CBF is the ratio of the cerebral perfusion pressure to cerebrovascular resistance. Therefore, physical factors that can change these two parameters can affect CBF. Moderate changes in the mean arterial blood pressure (MABP) have not been found to cause significant changes in CBF of infants with intact cerebral autoregulation [9]. If abrupt changes in the MABP are prevented, and in the absence of hypoglycemia and anemia that can increase CBF, the most important factors affecting the CBF of mechanically ventilated infants with intact CBF autoregulation are considered
n
Tel.: +1 714 281 5859; fax: +1 714 281 1360. E-mail address: [email protected]
0010-4825/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.compbiomed.2013.06.015
to be the arterial partial pressures of carbon dioxide and oxygen [7–9]. Therefore, proper regulation of infant's blood gases through appropriate selection of ventilator settings can significantly help to reduce hazardous disturbances of CBF that can lead to periventricular hemorrhage and hypoxic-ischemic brain injuries in premature infants. In this article, a decision support system (DSS) is presented by which the effects of different ventilator treatments on infants' CBF are simulated by using a mathematical model. At this time, continuous monitoring systems utilizing non-invasive technologies such as near-infrared spectroscopy and Doppler ultrasound are available that can be used to measure or estimate the changes in the infants' CBF [8,10]. However, routine clinical application of these systems is questionable due to their limitations and practical problems [8]. In the absence of direct CBF monitoring systems, or in conjunction with such support technologies, the technique described in this article can be used at bedside to help the intensivist choose appropriate ventilator settings for infants on mechanical ventilation and avoid hazardous disturbances in CBF and their untoward consequences. 2. Methodology A block diagram of the mathematical system used in the DSS is shown in Fig. 1. In this system, chemical control of infant's CBF is combined with a mathematical model of the infant's respiratory system [11,12] to simulate the effects of different ventilation treatments on the infant's CBF. The simulation results are
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
1403
Fig. 2. %ΔCBF as a function of P aCO2 .
Fig. 3. %ΔCBF as a function of P aO2 . Fig. 1. A block diagram of the computerized system to predict changes in the CBF of infants in response to ventilatory settings.
compared to reported clinical studies to assess the effectiveness of the system.
due to changes in P aCO2 . Combining Eqs. (2) and (3) yields
2.1. The CBF controller To design the CBF controller, experimental data from the literature were compiled to derive mathematical relations between the infants' CBF and arterial blood gases. In compiling and analyzing the experimental data, care was taken to choose the results obtained when both the arterial partial pressure of carbon dioxide, P aCO2 , and the arterial partial pressure of oxygen, P aO2 , were monitored at the same time. For the CBF response to P aCO2 , the results reported in several references were analyzed [13–16]. To measure the changes in CBF in response to P aCO2 , Doppler ultrasound techniques were used in References [13–15], and the intravenous Xe clearance technique was employed in the experiments of Reference [16]. The experimental data obtained for mechanically ventilated infants with gestational ages ranging from 26 to 33 weeks were selected. The median gestational age in this group was 29 weeks. For the CBF response to changes in P aO2 , the experimental data reported in References [15,17] were analyzed. In both studies, Doppler ultrasound techniques were used to measure CBF. Experimental data obtained for infants under 37 weeks of gestation were used for analysis. The median gestational age of these infants was 31 weeks ranging from 25 to 36 weeks. Percentage variations in CBF in comparison to normal CBF values were obtained and plotted as functions of P aCO2 and P aO2 . Normal CBF was considered at P aCO2 ¼40 mm Hg, and P aO2 ¼100 mm Hg. Figs. 2 and 3 show the percentage variations in infants' CBF (%ΔCBF) as compared to normal versus P aCO2 , and P aO2 respectively. Curve fitting techniques and linear regression analyses were performed on the data presented in Figs. 2 and 3. The results of the linear regression analyses are %ΔCBF ¼ 3:67ðP aCO2 –40Þ þ 2:257
due to changes in P aO2 (with r ¼0.90698).
2 %ΔCBF ¼ 0:0933½ΔP aO2 þ 3:69354½ΔP aCO2 0:00458 ΔP aCO2 þ 3:29354
ð4Þ
due to changes in blood gases where ΔP aO2 ¼ 100 P aO2 and ΔP aCO2 ¼P aCO2 40, and P aO2 and P aCO2 are in mm Hg. Eq. (4) gives the steady state value of %ΔCBF in relation to changes in arterial blood gases. In the controller, the steady state value of CBF is found by adding %ΔCBF from Eq. (4) to normal CBF, and the instantaneous value of CBF is determined by adding a 1st order lag with a time constant of 5 s. 2.2. The mathematical model of the infant's respiratory system The other building blocks of the system in Fig. 1 consisting of Lungs, Body Tissue, Transport Delay, and the Brain Tissue are parts of a mathematical model of the infant's respiratory system [12]. This detailed model of the human respiratory system has been used by a number of researchers to simulate the neonatal respiratory system e.g., [18–20]. The peripheral and central respiratory receptors and the respiratory controller of the model were not included in the system of Fig. 1 because in this DSS the respiratory drive signal is not provided by the infant's own respiratory control centers and is coming from the ventilator. The effects of an anatomical shunt in the lung is included in the system and the lung volume is time varying. The mathematical equations describing different building blocks of the model that have been incorporated in the DSS of Fig. 1 have been presented in Reference [12] and are not repeated here for brevity.
ð1Þ
3. Results
ð2Þ
Figs. 4–9 show a series of simulation results by using the presented DSS. In all the simulation tests, the patient is modeled as a premature infant of 1.8 kg weight, with respiratory distress syndrome (RDS), and a respiratory dynamic compliance of 2 ml/
due to changes in P aCO2 (with r ¼0.85398) and %ΔCBF ¼ 0:0933ð100P aO2 Þ þ 0:876
With regard to %ΔCBF response to P aCO2 , a quadratic fit was found to be more accurate than a linear formula. That quadratic equation was found as 2 %ΔCBF ¼ 2:41754 þ 3:69354ðP aCO2 40Þ0:00458 P aCO2 –40 ð3Þ
1404
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
Fig. 4. A set of simulation results of the DSS. The infant's ventilation and oxygenation are normal.
Fig. 6. A set of simulation results of the DSS in response to mild hyperventilation. The infant's oxygenation is normal.
Fig. 5. A set of simulation results of the DSS. The infant's oxygenation is normal while ventilation is low, resulting in hypercapnia.
Fig. 7. A set of simulation results of the DSS in response to low oxygen breathing.
cm H2O. A lung shunt ratio of 10% is assumed and the level of positive end-expiratory pressure (PEEP) is 5 cm H2O in all the tests.
Fig. 4a and b shows the simulation results with a respiratory frequency set at 40 breaths/min, minute ventilation at 0.59 l/min, and an inspired fraction of oxygen, F IO2 , set at 22.5%. Fig. 4a shows
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
1405
In Fig. 6, F IO2 is set at 22% and minute ventilation is raised to 0.68 l/min to induce hypocapnia. Fig. 7 shows the simulation results of the DSS with minute ventilation at 0.59 l/min, and F IO2 set at 18.5% to assess the effects of induced hypoxemia. Fig. 8 shows the simulation results of blood gases and CBF in response to induced hyperoxemia with minute ventilation at 0.59 l/min and F IO2 set at 28%. Fig. 9 shows the simulation results of the DSS with combined effects of hypoxemia and hypercapnia. In this test, minute ventilation is set at 0.5 l/min, and F IO2 is at the ambient air concentration of 21%.
4. Discussion
Fig. 8. A set of simulation results of the DSS in response to hyperoxia.
Fig. 9. A set of simulation results of the DSS in response to combined effects of hypoventilation and mild hypoxia.
the simulation results of the infant's P aO2 , and P aCO2 , while Fig. 4b shows the CBF predictions of the DSS. Fig. 5a and b shows the simulation results of blood gases and CBF for the same infant with F IO2 at 25%, and minute ventilation reduced to 0.5 l/min to induce hypercapnia.
The results of Fig. 4a shows that with the selected ventilatory settings, P aO2 rises close to 98 mm Hg and P aCO2 remains around 40 mm Hg over time. Fig. 4b shows the infant's CBF response which rises by about 9.5% initially due to a slight transient increase in P aCO2 and the initial low value of P aO2 . As shown in Fig. 4b, after an initial rise, CBF falls with time and approaches its normal level of 3.33 ml/s (corresponding to 200 ml/min or 11.1 ml/100 g/min) as blood gases approach and remain within normal limits. The results of Fig. 5a show an increase in P aCO2 over time to close to 50 mm Hg due to hypoventilation. The P aO2 value remains around 103 mm Hg during most of the time due to oxygen supplementation. Fig. 5b shows the CBF response due to induced hypercapnia. It is seen that CBF rises steadily to about 36% higher than its initial value to about 4.54 ml/s, corresponding to around 15 ml/100 g/min in 100 min of simulation in response to hypercapnia. The P aO2 level remains normal and does not cause any change in CBF in this test. These results are in good agreement with experimental findings [9,14,16]. Fig. 6a shows the simulation blood gas results of the DSS when the infant's ventilation is increased to 0.68 l/min to induce hypocapnia. The F IO2 level is kept near the ambient oxygen concentration at 22% to avoid hyperoxemia. As seen in Fig. 6a, P aO2 remains normal at about 100 mm Hg and P aCO2 falls to around 32.5 mm Hg over time as a result of hyperventilation. Fig. 6b shows the infant's CBF response to the induced hypocapnia. As seen, CBF falls to around 2.53 ml/s (151.8 ml/min corresponding to 8.43 ml/100 g/ min) with time. This 24% reduction in CBF in response to about 7.5 mm Hg drop in P aCO2 is in good agreement with experimental observations [16]. Fig. 7a shows the DSS's simulation results of the infant's blood gases with imposed low oxygen breathing. As seen, P aCO2 remains close to 40 mm Hg due to adequate ventilation settings but P aO2 falls to about 68.5 mm Hg with time due to low oxygen breathing. Fig. 7b shows the CBF response of the DSS in this test. As seen, there is an initial 11% rise in CBF to about 3.7 ml/s (222 ml/min or 12.3 ml/100 g/min) due to the initial rapid fall of P aO2 to about 65 mm Hg while P aCO2 rises to around 41.5 mm Hg. As P aO2 stabilizes around 68.5 mm Hg and P aCO2 decreases by a few mm Hg during the test, CBF approaches 3.45 ml/s that corresponds to 11.5 ml/100 g/min for this infant over time. This represents a final increase of about 3.6% in CBF due to the induced hypoxemia. These results show general agreement with the experimental data [17]. Fig. 8a shows the P aO2 and P aCO2 results of the DSS when oxygen supplementation provided to the infant is increased (F IO2 is 28%). This figure shows that P aCO2 remains around a normal value of 40 mm Hg while P aO2 rises to about 135 mm Hg over time. Fig. 8b shows that the infant's CBF falls to about 3.22 ml/s (i.e., 10.7 ml/ 100 g/min) in this test. This represents a fall of 3.3% in CBF due to the induced hyperoxemia. There is general agreement between these results and the experimental observations in preterm infants
1406
F.T. Tehrani / Computers in Biology and Medicine 43 (2013) 1402–1406
[17]. For normal infants however, the sensitivity of the CBF response to changes in P aO2 may be significantly higher than in premature infants as reported in the literature [17]. In Fig. 9a and b, the simulation results of the DSS are shown when ventilation is reduced to 0.5 l/min and no oxygen supplementation is provided to the infant (F IO2 is 21%). As a result of these ventilation settings, P aCO2 rises to about 48.5 mm Hg due to hypoventilation and P aO2 stabilizes around 75 mm Hg over time. Fig. 9b shows a steady increase in the infant's CBF response to about 4.6 ml/s (i.e., 15.3 ml/100 g/min) in response to combined effects of hypercapnia and hypoxemia. This represents an increase of about 38% in CBF. These results are in general agreement with experimental data for premature infants [9,14,16,17]. Once again, a significantly higher sensitivity in the CBF response to variations in P aO2 may be expected for normal full-term infants [17]. 5. Summary A computerized decision support system is presented for predicting the effects of different ventilatory settings on the cerebral blood flow of premature infants. The results of this study show that the presented system can be used to estimate the effects of changes in various ventilatory settings including minute ventilation, tidal volume, respiratory rate, PEEP, and F IO2 on the cerebral blood flow of preterm infants on mechanical ventilation. In the absence of direct monitoring systems, the system presented in this article can be used to estimate the cerebral blood flow variations for infants on mechanical ventilation provided the infants have intact cerebral autoregulation. Conflict of interest statement None declared.
[6] P. Sapieha, J.S. Joyal, J.C. Rivera, E. Kermorvant-Duchemin, F. Sennlaub, P. Hardy, P. Lachapelle, S. Chemtob, Retinopathy of prematurity: understanding ischemic retinal vasculopathies at an extreme of life, J. Clin. Invest. 120 (9) (2010) 3022–3032. [7] G. Greisen, Cerebral blood flow and energy metabolism in the newborn, Clin. Perinatol. 24 (3) (1997) 531–546. [8] K.D. Liem, G. Greisen, Monitoring of cerebral haemodynamics in newborn infants, Early Hum. Dev. 86 (3) (2010) 155–158. [9] J.R. Kaiser, G.H. Gauss, D.K. Williams, Tracheal suctioning is associated with prolonged disturbances of cerebral hemodynamics in very low birth weight infants, J. Perinatol. 28 (1) (2008) 34–41. [10] A. Milan, F. Freato, V. Vanzo, L. Chiandetti, P. Zaramella, Influence of ventilation mode on neonatal cerebral blood flow and volume, Early Hum. Dev. 85 (7) (2009) 415–419. [11] F.T. Tehrani, Autoregulation of the cerebral blood flow in preterm infants, in: Proceedings of IEEE-EMBC and CMBEC Conference, 1995, pp. 635–636. [12] F.T. Tehrani, Mathematical analysis and computer simulation of the respiratory system in the newborn infant, IEEE Trans. Biomed. Eng. 40 (5) (1993) 475–481. [13] M.I. Levene, D. Shortland, N. Gibson, D.H. Evans, Carbon dioxide reactivity of the cerebral circulation in extremely premature infants: effects of postnatal age and Indomethacin, Pediatr. Res. 24 (2) (1988) 175–179. [14] A.C. Fenton, K.L. Woods, D.H. Evans, M.I. Levene, Cerebrovascular carbon dioxide reactivity and failure of autoregulation in preterm infants, Arch. Dis. Child. 67 (7_Spec_No) (1992) 835–839. [15] S.T. Kempley, H.R. Gamsu, Arterial blood pressure and blood flow velocity in major cerebral and visceral arteries—I. Interindividual differences, Early Hum. Dev. 34 (3) (1993) 227–232. [16] O. Pryds, G. Greisen, Effects of P aCO2 and haemoglobin concentration on day to day variation of CBF in preterm neonates, Acta Paediatr. Scand. 360 (1989) 33–36. [17] S. Niijima, D.B. Shortland, M.I. Levene, D.H. Evans, Transient hyperoxia and cerebral blood flow velocity in infants born prematurely and at full term, Arch. Dis. Child. 63 (10_Spec_No) (1988) 1126–1130. [18] M.A. Mananas, C. Navarro, S. Romero, R. Grino, R. Rabinovich, S. Benito, P. Caminal, Control system response of different respiratory models under ventilatory stimuli and pathologies. in: Proceeding of the 15th World Congress of the International Federation on Automatic Control (IFAC), vol. 15, 2002, pp. 2317–2322. [19] R. Farrenkopf, Automatic Control of Oxygen Concentration to Premature Neonates. VPMA Progress Reports and Papers, 2002. Available from: 〈http:// www.docstoc.com/docs/95021462/ Automatic-Control-of-Oxygen-Flow-to-Premature-Neonates〉. [20] A.M. Hernandez, M.A. Mananas, R. Costa-Castello, Learning respiratory system function in BME studies by means of a virtual laboratory: respilab, IEEE Trans. Educ. 51 (1) (2008) 24–34.
References [1] T.A. Merritt, Respiratory distress, in: M. Ziai, T.A. Clarke, T.A. Merritt (Eds.), Assessment of the Newborn, a Guide for the Practitioner, 1st ed.,Little, Brown and Company, Boston, 1984. [2] N.C. Bradburn, Neurological causes of respiratory distress, in: R.L. Schreiner, N.C. Bradburn (Eds.), Care of the Newborn, 2nd ed.,Raven Press, New York, 1988. [3] S. Farzan, Respiratory distress syndrome of the newborn: hyaline membrane disease, in: S. Farzan (Ed.), A Concise Handbook of Respiratory Distress, 2nd ed.,Reston Publishing Company Inc., A Prentice Hall Company Reston, 1985. [4] D.L. Levin, Hyaline membrane disease, in: D.L. Levin, F.C. Morriss, G.C. Moore (Eds.), A Practical Guide to Pediatric Intensive Care, The C.V. Mosby Company St. Louis, 1979. [5] F. Riedel, Long term effects of artificial ventilation in neonates, Acta Paediatr. Scand. 76 (1) (1987) 24–29.
Fleur T. Tehrani holds a B.S. degree in electrical engineering from Sharif University of Technology, Tehran, Iran, a graduate diploma from Imperial College of Science and Technology, London, and M.S. and Ph.D. degrees in electrical engineering from the University of London, UK. She has been on the faculty of California State University, Fullerton, since 1985 where she is currently a tenured professor of electrical engineering. She is the inventor of metabolic rate monitors and several respiratory assist devices including an advanced respiratory control system known as “Adaptive Support Ventilation.” Dr. Tehrani is a Fellow of the Institution of Engineering and Technology (the IET), a Fellow of the Institute for the Advancement of Engineering, a senior member of the Institute of Electrical and Electronics Engineers (IEEE), and a National Life Member of Graduate Women in Science (GWIS).