Top Banner
Cerebral Monitoring in Cardiac Surgery: Sensitivity of Hb Co-variations and Individual Components Sergio A. Ramirez 1,2 , MD, Harry L. Graber 1 , PhD, LeRone Simpson 1,2 , MD, Yong Xu 1 , PhD, Yaling Pei 1 , PhD, Douglas Pfeil 1 , BS, Vinay Tak 1,3 , MD, Joshua Burack 1,2 , MD, Wilson Ko 1 , MD, Randall L. Barbour 1 , PhD, and Daniel C. Lee 1,3 , MD 1 State University of New York, Downstate Medical Center, Box 40, 450 Clarkson Avenue, Brooklyn, New York 11203, 2 The Brooklyn Hospital Center, Brooklyn NY 11201, 3 Interfaith Medical Center, Brooklyn, NY 11213 Discussion The percentage of event-sensitive SM-based c-v metrics is significantly greater than the corresponding percentage of SM-based conventional metrics (p = 4.3×10 -9 , Fig. 3). While the overall number of S-M c-v metrics is larger than the corresponding number of conventional metrics by a factor of 6 (Table 1), the number of event- sensitive cv-state metrics is greater than its conventional counterpart by a factor of 18. The reason for our initial focus on SM-based metric is that these correspond to the only type of information that is obtained from FDA-approved devices comprising a single illumination site and a small number of detectors. In contrast, both c-v analysis and the computation of SSDs require spatially resolved images as input. The requisite real-time image reconstruction capability has been demonstrated previously. 7,8 Thus optical tomography has an important advantage over simple cerebral oximetry, which does not provide the spatial variability information necessary for image reconstruction. Nearly half of the conventional-analysis metrics show very little sensitivity to the occurrence of clinically significant events (Fig. 4). The other half initially seem to perform comparably to the most sensitive c-v metrics. However, only c-v metrics (Fig. 5) simultaneously yield potentially useful levels of sensitivity and specificity. In conclusion, our data provide credible preliminary evidence that the metrics obtained from analysis of the Hb image time series based on c-v states would be a more successful basis for an intraoperative monitoring tool, compared to evaluation of the same image data in terms of the conventional individual Hb signals (i.e., Hb oxy , Hb deoxy , Hb total ). References 1 M. Wolf et al., “Progress of near-infrared spectroscopy and topography for brain and muscle clinical applications,” Journal of Biomedical Optics 12(6), 062104, (2007). 2 B.R. White et al., “Resting-state functional connectivity in the human brain revealed with diffuse optical tomography,” NeuroImage 47(1), 148-156, (2009). 3 G.W. Wylie et al., “Using co-variations in the Hb signal to detect visual activation: A near infrared spectroscopic imaging study,” NeuroImage 47(2), 473-481, (2009). 4 Y. Pei et al., “Influence of systematic errors in reference states on image quality and on stability of derived information for DC optical imaging,” Applied Optics 40(31), 5755-5769, (2001). 5 J.A. Hanley and B.J. McNeil, “The meaning and use of the area under the Receiver Operating Characteristic (ROC) curve,” Radiology 143(1), 29-36, (1982). 6 R.G Newcombe, “Statistical applications in orthodontics. Part 2: Confidence intervals for proportions and their differences,” Journal of Orthodontics 27, 339- 340, (2000). 7 C.H. Schmitz et al., “Instrumentation for Real-Time Dynamic Optical Tomography,” Photon Migration, Optical Coherence Tomography, and Microscopy (Proceedings of SPIE, Vol. 4431), Stefan Andersson-Engels, Michael Kaschke, Eds., pp. 282-291, (2001). 8 Y. Pei et al., “SVD-based normalized-transformed scheme for real-time DC optical tomography,” OSA Biomedical Topical Meetings, Advances in Optical Imaging and Photon Migration (Miami Beach, FL, April 7-10, 2002), pp. 629-631, (2002). Acknowledgements This work was supported under grants no. 5R42NS050007-03 and no. 5R44NS049734-03 to Randall L. Barbour; and by the New York State Department of Health. Introduction Near-infrared spectroscopy (NIRS) monitoring of cerebral oxygenation has been well described and utilized clinically. 1 In an accompanying submission, we demonstrated that devices having only a small number of optodes are unlikely to provide reliable representations of cerebral perfusion or to detect hemodynamically significant events, underscoring the merit of adopting high-density optode arrays. In this report we examine the strategies used for analysis and interpretation of data obtained by NIRS tomography with high-density optode arrays. We compared methods of examining hemoglobin (Hb) signals individually, 2 versus exploring the spatial and temporal components of the co-variations among oxygenated (Hb oxy ), deoxygenated (Hb deoxy ) and total (Hb total ) Hb, 3 in terms of their effectiveness in detecting the cerebral hemodynamic responses associated with intraoperative clinically significant changes in blood pressure. We present evidence documenting that the considered co-variational (c-v) analysis approach provides an enhanced capability, with a substantially reduced rate of false positive results, for detection of clinically significant hemodynamic events, compared to similar analyses based on the individual Hb signals (i.e., the standard approach). Table 1. Clinical information for the study participants. Abbreviations: AVR = aortic valve replacement, CABG = coronary artery bypass graft, CPB = cardiopulmonary bypass, VF = ventricular fibrillation. Retrograde Autologous Priming of CPB 4h 44m AVR – On-pump 27 F 6 Low CPB flow rate 3h 27m GABG – On-pump 58 F 5 Initiation of CPB 4h 32m GABG – On-pump 58 F 4 Cardiac Cannulation 3h 00m AVR – On-pump 67 M 3 VF / Cardioversion 5h 38m GABG – Off-pump 73 F 2 VF arrest / crash on CPB 5h 18m CABG – Off-pump converted to on-pump beating heart 65 F 1 Hemodynamic Event Measurement Duartion Surgery Age Gender Subject Methods Two-wavelength time series data were collected (NIRx DYNOTcompact) from six patients during heart surgery (Table 1), using the multi-channel optode arrays shown in Fig. 1. Previously described methods 4 were used to reconstruct three-dimensional images of the Hb states: (Hb oxy , Hb deoxy and Hb total ). Using the definitions indicated in Table 2, each image pixel was assigned to one of six co-variation (c-v) states that correspond to different conditions of O 2 supply/demand balance, for each image time frame. 3 Additional information that the c-v decomposition make available is illustrated in Fig. 2, using data from a forearm imaging study. The Hb and c-v state data were subsequently processed to yield multiple image metrics that involve elements of the spatial and temporal information domains derived from time periods corresponding either to clinically significant hemodynamic events or to control (i.e., clinically unimportant) intervals. A clinically significant hemodynamic event was defined as a change in mean arterial pressure (MAP) >20 mm Hg, with the episode lasting more than 2 min, and a control period by a change in MAP < 5 mm Hg. The metrics examined were the spatial mean (SM) value and spatial standard deviation (SSD) of the Hb oxy , Hb deoxy , Hb total and Hb O 2 saturation (O 2 Sat) time series, as well as the time-varying percentages of image volume (Vol%) occupied by each of the six c-v states. A total of 248 time-dependent metrics were computed; 216 of these correspond to metrics derived from c-v state analysis, while the remaining 32 correspond to metrics obtained from analysis of individual Hb states (conventional approach); at least one interval of each type (minimum = 1, maximum = 7) was chosen for each subject. Receiver Operating Characteristic analysis 5 was performed to determine which metrics can distinguish between control and event time intervals (true positive results), and which reveal differences between pairs of control intervals (false positive results). The number of metrics with a statistically significant ability (α = 0.05) to discriminates between segments was recorded, using the “score method” 6 to estimate the confidence intervals. Figure 1. a) Photograph of headgear with the four-site optode array; b) Diagram of the optode arrangements in each site, and their inter-site distances. 7 8 1 2 3 4 5 6 1 6 8 2 3 4 5 7 1 8 2 3 4 5 6 7 2 5 1 8 3 4 6 7 Site 1 Site 3 Site 4 Site 2 a b c a = 10.5 cm b = 6.5 cm c = 4.5 cm 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 1 6 8 2 3 4 5 7 1 6 8 2 3 4 5 7 1 8 2 3 4 5 6 7 1 8 2 3 4 5 6 7 2 5 1 8 3 4 6 7 2 5 1 8 3 4 6 7 Site 1 Site 3 Site 4 Site 2 a b c a = 10.5 cm b = 6.5 cm c = 4.5 cm b) a) Table 2. Concentration changes relative to a temporal mean value, associated with Hb levels influenced by autoregulation. Comp. O 2 excess Uncomp. O 2 excess Balanced Comp. O 2 debt Uncomp. O 2 debt Balanced + + + Hb total + + + Hb deoxy + + + Hb oxy State 6 State 5 State 4 State 3 State 2 State 1 Hemoglobin State Results Figure 3. For each of the 17 clinically significant event time intervals: number of spatial mean- based metrics, expressed as percentage of theoretical maximum (i.e., 16 for the integrated Hb signals, and 96 for the c-v states), with statistically significant (α = 0.05) ability to distinguish the clinically significant and baseline time intervals. Also shown are the Mean and STD across all events. 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 STD Mean cv - states conventional 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 0 6 12 18 24 29 35 41 47 53 59 65 71 76 82 88 94 100 % of Metrics Sensitivity (%) cv - states conventional Figure 4. Diagnostic sensitivity histograms for the complete sets of conventional and c-v state-based metrics. Bar heights indicate the percentage of metrics that have the sensitivity values shown on the horizontal axis. The c-v metric distribution (min. = 12%, max. = 94%, avg. = 49%) is shifted rightward by a small amount in comparison to the conventional metric distribution (min. = 0%, max. = 88%, avg. = 46%) . 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 0.0 14.3 28.6 42.9 57.1 71.4 85.7 100.0 Specificity (%) % of Metrics (Sensitivity 70%) cv - states conventional Figure 5. Diagnostic specificity histograms, for all conventional (n = 12) and c-v (n = 35) metrics having 70% sensitivity. Only one of the conventional metrics has specificity > 50%, and all are < 60%. In contrast, several c-v metrics have specificity > 70%, thereby meriting further study. 1 2 3 Hb oxy Hb deoxy Hb total Figure 2. (a) Conventional reconstructed images of Hb oxy, Hb deoxy, Hb total , in an 2D section through a volunteer’s forearm, at a time frame representative of the entire baseline time series. All three images are qualitatively similar, with superficial veins dominating the signal in every case. Deeper lying dynamic features are present but obscured by the larger-amplitude superficial information. (b) Columns 1-3: The same image data as in (a), but fractionated according to which one of the 6 c-v states (Table 2) is present in each pixel during the selected time frame. (That is, each pixel is set to a value of 0 if the c-v state is different from the one indicated in the row heading.) In this representation, features that were obscured in conventional analysis are unveiled. Furthermore, anatomical features can be tentatively assigned a physiological significance, on the basis of which state reveals its presence. Column 4: time fraction images for each of the 6 c-v states. These spatial maps show how much time each pixel spends in a particular c-v state during the full course of the baseline time interval (approximately 500 time frames). This analysis is informative because it is amplitude- insensitive, provides a way of summarizing the information in any number of time frames with a small fixed number of spatial maps, and provides information on which tissue regions principally experience the physiological phenomena associated with each c- v state. State Hb-oxy Hb-deoxy Hb-total Time Fraction 1 2 3 4 5 6 a b
1

Cerebral Monitoring in Cardiac Surgery: Sensitivity …...Cerebral Monitoring in Cardiac Surgery: Sensitivity of Hb Co-variations and Individual Components Sergio A. Ramirez 1,2 ,

Jun 05, 2020

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Cerebral Monitoring in Cardiac Surgery: Sensitivity …...Cerebral Monitoring in Cardiac Surgery: Sensitivity of Hb Co-variations and Individual Components Sergio A. Ramirez 1,2 ,

% S

patia

l Mea

n M

etric

s

Clinically Significant Time Periods

Cerebral Monitoring in Cardiac Surgery: Sensitivity of Hb Co-variations and Individual Components

Sergio A. Ramirez1,2, MD, Harry L. Graber1, PhD, LeRone Simpson1,2, MD, Yong Xu1, PhD,Yaling Pei1, PhD, Douglas Pfeil1, BS, Vinay Tak1,3, MD, Joshua Burack1,2, MD, Wilson Ko1, MD,

Randall L. Barbour1, PhD, and Daniel C. Lee1,3, MD1State University of New York, Downstate Medical Center, Box 40, 450 Clarkson Avenue, Brooklyn, New York 11203,

2The Brooklyn Hospital Center, Brooklyn NY 11201, 3Interfaith Medical Center, Brooklyn, NY 11213

Discussion• The percentage of event-sensitive SM-based c-v metrics is significantly greater

than the corresponding percentage of SM-based conventional metrics (p = 4.3×10-9, Fig. 3).• While the overall number of S-M c-v metrics is larger than the corresponding

number of conventional metrics by a factor of 6 (Table 1), the number of event-sensitive cv-state metrics is greater than its conventional counterpart by a factor of 18.

• The reason for our initial focus on SM-based metric is that these correspond to the only type of information that is obtained from FDA-approved devices comprising a single illumination site and a small number of detectors.

• In contrast, both c-v analysis and the computation of SSDs require spatially resolved images as input.• The requisite real-time image reconstruction capability has been

demonstrated previously.7,8

• Thus optical tomography has an important advantage over simple cerebral oximetry, which does not provide the spatial variability information necessary for image reconstruction.

• Nearly half of the conventional-analysis metrics show very little sensitivity to the occurrence of clinically significant events (Fig. 4). The other half initially seem toperform comparably to the most sensitive c-v metrics.• However, only c-v metrics (Fig. 5) simultaneously yield potentially useful levels

of sensitivity and specificity.• In conclusion, our data provide credible preliminary evidence that the metrics

obtained from analysis of the Hb image time series based on c-v states would be a more successful basis for an intraoperative monitoring tool, compared to evaluation of the same image data in terms of the conventional individual Hbsignals (i.e., Hboxy, Hbdeoxy, Hbtotal).

References1 M. Wolf et al., “Progress of near-infrared spectroscopy and topography for brain

and muscle clinical applications,” Journal of Biomedical Optics 12(6), 062104, (2007).

2 B.R. White et al., “Resting-state functional connectivity in the human brain revealed with diffuse optical tomography,” NeuroImage 47(1), 148-156, (2009).

3 G.W. Wylie et al., “Using co-variations in the Hb signal to detect visual activation: A near infrared spectroscopic imaging study,” NeuroImage 47(2), 473-481, (2009).

4 Y. Pei et al., “Influence of systematic errors in reference states on image quality and on stability of derived information for DC optical imaging,” Applied Optics40(31), 5755-5769, (2001).

5 J.A. Hanley and B.J. McNeil, “The meaning and use of the area under the Receiver Operating Characteristic (ROC) curve,” Radiology 143(1), 29-36, (1982).

6 R.G Newcombe, “Statistical applications in orthodontics. Part 2: Confidence intervals for proportions and their differences,” Journal of Orthodontics 27, 339-340, (2000).

7 C.H. Schmitz et al., “Instrumentation for Real-Time Dynamic Optical Tomography,” Photon Migration, Optical Coherence Tomography, and Microscopy (Proceedings of SPIE, Vol. 4431), Stefan Andersson-Engels, Michael Kaschke, Eds., pp. 282-291, (2001).

8 Y. Pei et al., “SVD-based normalized-transformed scheme for real-time DC optical tomography,” OSA Biomedical Topical Meetings, Advances in Optical Imaging and Photon Migration (Miami Beach, FL, April 7-10, 2002), pp. 629-631, (2002).

AcknowledgementsThis work was supported under grants no. 5R42NS050007-03 and no. 5R44NS049734-03 to Randall L. Barbour; and by the New York State Department of Health.

Introduction • Near-infrared spectroscopy (NIRS) monitoring of cerebral oxygenation has been

well described and utilized clinically.1• In an accompanying submission, we demonstrated that devices having only a

small number of optodes are unlikely to provide reliable representations of cerebral perfusion or to detect hemodynamically significant events, underscoring the merit of adopting high-density optode arrays.

• In this report we examine the strategies used for analysis and interpretation of data obtained by NIRS tomography with high-density optode arrays.

• We compared methods of examining hemoglobin (Hb) signals individually,2 versus exploring the spatial and temporal components of the co-variations among oxygenated (Hboxy), deoxygenated (Hbdeoxy) and total (Hbtotal) Hb,3 in terms of their effectiveness in detecting the cerebral hemodynamic responses associated with intraoperative clinically significant changes in blood pressure.

• We present evidence documenting that the considered co-variational (c-v) analysis approach provides an enhanced capability, with a substantially reduced rate of false positive results, for detection of clinically significant hemodynamic events, compared to similar analyses based on the individual Hb signals (i.e., the standard approach).

Table 1. Clinical information for the study participants. Abbreviations: AVR = aortic valve replacement, CABG = coronary artery bypass graft, CPB = cardiopulmonary bypass, VF = ventricular fibrillation.

Retrograde AutologousPriming of CPB4h 44mAVR – On-pump27F6

Low CPB flow rate3h 27mGABG – On-pump58F5

Initiation of CPB4h 32mGABG – On-pump58F4

Cardiac Cannulation3h 00mAVR – On-pump67M3

VF / Cardioversion5h 38mGABG – Off-pump73F2

VF arrest / crash on CPB5h 18mCABG – Off-pump

converted to on-pump beating heart

65F1

Hemodynamic EventMeasurement DuartionSurgeryAgeGenderSubject

Retrograde AutologousPriming of CPB4h 44mAVR – On-pump27F6

Low CPB flow rate3h 27mGABG – On-pump58F5

Initiation of CPB4h 32mGABG – On-pump58F4

Cardiac Cannulation3h 00mAVR – On-pump67M3

VF / Cardioversion5h 38mGABG – Off-pump73F2

VF arrest / crash on CPB5h 18mCABG – Off-pump

converted to on-pump beating heart

65F1

Hemodynamic EventMeasurement DuartionSurgeryAgeGenderSubject

Methods• Two-wavelength time series data were collected (NIRx DYNOTcompact) from six patients during heart surgery (Table

1), using the multi-channel optode arrays shown in Fig. 1.• Previously described methods4 were used to reconstruct three-dimensional images of the Hb states: (Hboxy, Hbdeoxy

and Hbtotal). • Using the definitions indicated in Table 2, each image pixel was assigned to one of six co-variation (c-v) states that

correspond to different conditions of O2 supply/demand balance, for each image time frame.3• Additional information that the c-v decomposition make available is illustrated in Fig. 2, using data from a forearm

imaging study.• The Hb and c-v state data were subsequently processed to yield multiple image metrics that involve elements of the

spatial and temporal information domains derived from time periods corresponding either to clinically significant hemodynamic events or to control (i.e., clinically unimportant) intervals.• A clinically significant hemodynamic event was defined as a change in mean arterial pressure (MAP) >20 mm Hg,

with the episode lasting more than 2 min, and a control period by a change in MAP < 5 mm Hg.• The metrics examined were the spatial mean (SM) value and spatial standard deviation (SSD) of the Hboxy,

Hbdeoxy, Hbtotal and Hb O2 saturation (O2Sat) time series, as well as the time-varying percentages of image volume (Vol%) occupied by each of the six c-v states.

• A total of 248 time-dependent metrics were computed; 216 of these correspond to metrics derived from c-v state analysis, while the remaining 32 correspond to metrics obtained from analysis of individual Hb states (conventional approach); at least one interval of each type (minimum = 1, maximum = 7) was chosen for each subject.

• Receiver Operating Characteristic analysis5 was performed to determine which metrics can distinguish between control and event time intervals (true positive results), and which reveal differences between pairs of control intervals (false positive results).

• The number of metrics with a statistically significant ability (α = 0.05) to discriminates between segments was recorded, using the “score method”6 to estimate the confidence intervals.

Figure 1. a) Photograph of headgear with the four-site optodearray; b) Diagram of the optode arrangements in each site, and their inter-site distances.

78 1234

561

68

2 3 4

57

18

234

5672

5

1 83 4

6 7Site 1

Site 3 Site 4

Site 2

a

bc

a = 10.5 cm b = 6.5 cm c = 4.5 cm

78 1234

5678 1234

561

68

2 3 4

57

16

82 3 4

57

18

234

5671

8234

5672

5

1 83 4

6 72

5

1 83 4

6 7Site 1

Site 3 Site 4

Site 2

a

bc

a = 10.5 cm b = 6.5 cm c = 4.5 cmb)a)

Table 2. Concentration changes relative to a temporal mean value, associated with Hb levels influenced by autoregulation.

Comp. O2 excess

Uncomp. O2 excessBalancedComp.

O2 debtUncomp. O2 debtBalanced

–+++––Hbtotal

––+++–Hbdeoxy

+++–––Hboxy

State 6

State 5

State 4

State 3

State 2

State 1

Hemoglobin State

Comp. O2 excess

Uncomp. O2 excessBalancedComp.

O2 debtUncomp. O2 debtBalanced

–+++––Hbtotal

––+++–Hbdeoxy

+++–––Hboxy

State 6

State 5

State 4

State 3

State 2

State 1

Hemoglobin State

Results

Figure 3. For each of the 17 clinically significant event time intervals: number of spatial mean-based metrics, expressed as percentage of theoretical maximum (i.e., 16 for the integrated Hb signals, and 96 for the c-v states), with statistically significant (α = 0.05) ability to distinguish the clinically significant and baseline time intervals. Also shown are the Mean and STD across all events.

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

STD

Mea

n

cv - statesconventional

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

0 6 12 18 24 29 35 41 47 53 59 65 71 76 82 88 94 100

% o

f M

etric

s

Sensitivity (%)

cv - statesconventional

Figure 4. Diagnostic sensitivity histograms for the complete sets of conventional and c-v state-based metrics. Bar heights indicate the percentage of metrics that have the sensitivity values shownon the horizontal axis. The c-v metric distribution (min. = 12%, max. = 94%, avg. = 49%) is shifted rightward by a small amount in comparison to the conventional metric distribution (min. = 0%, max. = 88%, avg. = 46%) .

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

0.0 14.3 28.6 42.9 57.1 71.4 85.7 100.0Specificity (%)

% o

f Met

rics

(Sen

sitiv

ity ≥

70%

)

cv - statesconventional

Figure 5. Diagnostic specificity histograms, for all conventional (n = 12) and c-v (n = 35) metrics having ≥ 70% sensitivity. Only one of the conventional metrics has specificity > 50%, and all are < 60%. In contrast, several c-v metrics have specificity > 70%, thereby meriting further study.

1 2 3

Hboxy Hbdeoxy Hbtotal

Figure 2. (a) Conventional reconstructed images of Hboxy,Hbdeoxy, Hbtotal, in an 2D section through a volunteer’s forearm, at a time frame representative of the entire baseline time series. All three images are qualitatively similar, with superficial veins dominating the signal in every case. Deeper lying dynamic features are present but obscured by the larger-amplitude superficial information. (b) Columns 1-3: The same image data as in (a), but fractionated according to which one of the 6 c-vstates (Table 2) is present in each pixel during the selected time frame. (That is, each pixel is set to a value of 0 if the c-v state is different from the one indicated in the row heading.) In this representation, features that were obscured in conventional analysis are unveiled. Furthermore, anatomical features can be tentatively assigned a physiological significance, on the basis of which state reveals its presence. Column 4: time fraction images for each of the 6 c-v states. These spatial maps show how much time each pixel spends in a particular c-v state during the full course of the baseline time interval (approximately 500 time frames). This analysis is informative because it is amplitude-insensitive, provides a way of summarizing the information in any number of time frames with a small fixed number of spatial maps,and provides information on which tissue regions principally experience the physiological phenomena associated with each c-v state.

State HbOxy HbDeoxy HbTotal Time Fraction

1

2

3

4

5

6

State Hb-oxy Hb-deoxy Hb-total Time Fraction

1

2

3

4

5

6

State HbOxy HbDeoxy HbTotal Time Fraction

1

2

3

4

5

6

State Hb-oxy Hb-deoxy Hb-total Time Fraction

1

2

3

4

5

6

State HbOxy HbDeoxy HbTotal Time Fraction

1

2

3

4

5

6

State Hb-oxy Hb-deoxy Hb-total Time Fraction

1

2

3

4

5

6

State HbOxy HbDeoxy HbTotal Time Fraction

1

2

3

4

5

6

State HbOxy HbDeoxy HbTotal Time Fraction

1

2

3

4

5

6

State Hb-oxy Hb-deoxy Hb-total Time Fraction

1

2

3

4

5

6

a b