Validation of real-time continuous perfusion ... - Springer Link

f

Validation of real-time continuous perfusion measurement G. T. Martin I

H.F. B o w m a n 2

1Thermal Technologies, Inc. Cambridge, MA 21-1arvard-MIT Division of Health Sciences and Technology

Abstract--Perfusion, the rate at which blood in tissue is replenished at the capillary level, is a p r i m a r y factor in the transport of heat, drugs, oxygen and nutrients. While there have been m a n y measurement techniques proposed, most do not lend themselves to routine, continuous and real-time use. A m i n i m a l l y invasive probe, called the thermal diffusion probe (TDP), which uses a self-heated thermistor to measure absolute perfusion continuously and in real time, was validated at l o w flows with the microsphere technique. In 27 rabbits, simultaneous TDP measurements were made in liver from 0 to 6 0 m l m i n 1 lOOg 1. The TDP perfusion correlated well with the microspheres (R2= 0.898) and the agreement between techniques is very good with a slope close to unity (0.921) and an intercept close to zero (0.566mlmin 1 lOOg 1). Variability between the two techniques was p r i m a r i l y due to the sampling error from the microsphere "snap shot" of periodic blood flow when compared with the continuous TDP perfusion measurement. The ability to quantify local perfusion continuously and in real time m a y have a profound impact on patient management in a number of clinical areas such as organ transplantation, neurosurgery, oncology and others, in which quantitative knowledge of perfusion is of value. K e y w o r d s - - B l o o d flow, Perfusion, Thermal diffusion probe, Microspheres, Rabbits Med. Biol. Eng. Comput., 2000, 38, 319-325

J

1 Introduction

PERFUSION, TISSUE blood flow, can be defined as the rate at which the quantity of blood in a given mass or volume of tissue is replenished at the level of the capillary network (BOWMAN, 1984). Perfusion is a primary factor in the local transport of heat, drugs, oxygen, nutrients and waste products. The ability to measure perfusion in small volumes of tissue has been sought for many years, as this fundamental parameter holds the key to an improved understanding of both normal and pathologic physiology as well as the diagnosis and subsequent management of numerous medical problems. The ideal per fusion monitoring system would be able to routinely and continuously quantify perfusion in the region of interest, non-invasively or at least minimally invasively, in real time. The overall importance of tissue blood flow to human health has motivated the development of many measurement techniques, most of which, generally, do not lend themselves to routine, continuous and real-time measurement of per fusion. Such techniques include: radioactive tracer washout techniques; positron emission tomography (PET); magnetic resonance imaging (MRI); radioactive microspheres; and laser-Doppler flowmetry (LDF). PET uses short half-life isotopes and requires a special facility, while MRI may hold future promise. In any case, both of these techniques are complex, expensive and do not

Correspondence should be addressed to Dr G. T. Martin; emaih [email protected] First received 25 October 1999and in final form 14 February 2000 © IFMBE:2000

Medical & Biological Engineering & Computing 2000, Vol. 38

permit routine monitoring. Radioactive microspheres are suitable for perfusion validation studies in animal models, but are not appropriate for clinical measurements. Laser Doppler flowmetry (LDF) assesses the average tissue blood flow in a finite tissue volume continuously and in real time. A minimally invasive fibre optic probe carries coherent light to the tissue and the Doppler shift from the net erythrocyte velocity within the laser-illuminated volume is measured. The LDF signal depends on a number of factors, such as hematocrit, red blood cell velocity, vascular geometry and tissue optical properties which vary according to tissue type (SMITS et al., 1986). Thus, it is not appropriate to apply the LDF calibration and measurements from one tissue type to another tissue type. Furthermore, it is unlikely that LDF units can be converted to absolute blood flow. LDF may be useful, however, in correlating per fusion variation with other measurement techniques. Thermal clearance (also called thermal diffusion) has in the past been applied to the measurement of perfusion. CHOKSEY (1996) details part of the very long history of the development of thermal techniques for blood flow measurement. The main obstacle to the acceptance of thermal clearance methods, cited by CHOKSEY (1996) is the determination of the relationship between the thermal clearance (tissue heat transfer ability) and the blood flow. In contrast, BOWMAN (1984) reviewed many of the thermal-based techniques that use the tissue energy balance and a model of probe-tissue heat transfer to determine absolute tissue perfusion. One such technique to measure absolute flow is described by VALVANO et al. (1984a), who use a minimally invasive thermistor probe. This probe contains a thermistor embedded at the tip of a flexible catheter. The surface of the thermistor is controlled to a small increment ( ~ 2 :C) above the tissue base319

line temperature. The power dissipated in the thermistor (510 mW) provides a measure of the ability of the tissue to carry heat by both thermal conduction within the tissue and thermal convection due to tissue blood flow. In comparison to the radioactive microsphere technique, the authors found that the technique could measure absolute perfusion to an accuracy of about 10% over the range 60-200 ml mill -1 100 g-1 in an ex vivo rat liver model (VALVANO et al., 1984b). While these results are promising, such flow rates are typical of major organs under normal physiologic conditions. As a diagnostic technique, the full potential of per fusion quantification can only be realised when low pathophysiologic flows can be distinguished from the normal flow levels in healthy tissue. We have been working for some years on the development of thermal techniques for the measurement of tissue blood flow. The aim of this effort is to develop a minimally invasive probe that can continuously measure absolute tissue perfusion, in real time, with a resolution and accuracy better than 1 mlmin -1 100 g-1. Like VALVANO et al. (1984a) we model the coupled thermistor-tissue heat transfer and solve for the temperature distribution within the thermistor material and for the temperature distribution in the surrounding tissue, according to the Pennes 'bioheat' transfer equation (PENNES, 1948). In contrast to VALVANO et al. (1984a) we add a proximal thermistor to passively monitor and compensate for temporal tissue baseline temperature changes and we solve the tissue heat transfer equations continuously to provide perfusion in real time. The purpose of this paper is to report recent validation results in the measurement o f per fusion below 60 ml m i n - 1 100 g - 1 using the thermal diffusion probe (TDP).

2 M a t e r i a l s and m e t h o d s

Validation studies were carried out in a rabbit model to compare liver perfusion measured with the TDP to the 'gold standard' radioactive microsphere technique. All the experiments were carried out under institutional approval and conformed to the guidelines of the Guide f o r the Care and Use o f Laborato O, Animals, US National Institutes of Health (NIH) Publication 85-23, Revised 1985.

measurement. Thus to obtain a valid measure of total liver blood flow from the microsphere technique, the portal vein was clamped and the liver was supplied only by the hepatic artery which can carry microspheres. After the liver was surgically exposed, the TDP was placed 1.5 cm deep into the largest liver lobe and sutured in place. To maintain a thermally stable environment, the incision in the abdomen was closed. When thermal stability between the probe and the tissue was reached (in about 2 min), nearly continuous perfusion measurements were made throughout the experiment. TDP perfusion measurements taken immediately after microsphere injection were used for the comparison.

2.2 TDP perfitsion measurement The TDP perfusion measurements were taken using a TDP200* instrument. This high precision device excites the active thermistor to a constant temperature slightly above the tissue baseline (selectable at about 2 :C with a 0.001 :C stability); it collects data on the power dissipated in the active, heated thermistor and it constantly monitors the baseline tissue temperature using the additional passive thermistor placed outside the heated field. The time constant of the perfusion sensor response is about 0.1 s. Fig. 1 shows a schematic diagram of the probe. The ability to monitor baseline tissue temperature during data collection makes the perfusion measurement less prone to error from tissue thermal variations over extended periods of time. The control of the data collection, the A / D conversion, and the communication with the host computer are all done with an on-board microprocessor'S. The host computer collects data on the power dissipation in the active thermistor and solves the probe-tissue heat transfer equations to quantify perfusion in real time. This technique quantifies perfusion in a tissue volume of approximately 0.3 ml without a no-flow calibration measurement. This is achieved by determining the conductive properties of the tissue from the initial rate of propagation of the thermal field when the heat to the active thermistor is first turned on, and using the measure of thermal

I~ 5 mm passivet~rmistor /~"~'~~

~'N~

2.1 Surgical preparation

New Zealand White male rabbits* were given an intramuscular (IM) injection of ketamine/xylazine (35/5 m g k g -1 body weight) to sedate each animal prior to intubation. Anaesthesia was maintained with 1-2% isofluorane, administered via an anaesthesia apparatus'~. Ventilator settings were adjusted to maintain normal pH, pCO2, arterial pOe levels which were periodically monitored with arterial blood gas measurements throughout the experiment. A catheter was placed in a common carotid artery for pressure measurement and two other catheters were placed in the left and right femoral arteries for the withdrawal of blood samples for blood gas analysis and for microsphere sampling. The heart was exposed through a left thoracotomy and supported in a pericardial cradle so that a fourth catheter could be inserted into the left atrium to inject microspheres. The liver was exposed with a midline laparotomy to allow easy insertion of the TDP and the intact portal vein was closed with an aneurysm clamp. Since the portal vein supplies blood to the liver that has been drained through the gastrointestinal capillaries, this blood would not be able to carry the 15-pm microspheres to the liver tissue for blood flow

*Millbrook Farms, Amherst, MA tBoyle Model 50, Harris-Lake, Inc., Cleveland, OH 320

'm

) k

"J activethermistor, , ~ ,~

~

/i

~ . _ measurement field

[

[

~4

mm

Fig. 1 Schematic diagram o f the thermal diffitsion probe (TDP) showing the active, heated thermistor at the probe tip which produces a thermal measurement field in the sumvunding tissue. The size o f the thermal measurement field is dependent oil the tissue thermal properties" and the pelfusion: high perfusion produces a smaller thermal field. The diameter of the field is applvximately 4 mm for O.'pical values o f thermal properties" and perfusion. The passive thermistol; mounted 5 mm proximal to the probe tip, monitors the tissue baseline temperature variations

*Thermal Technologies, Inc., Cambridge, MA ~lntel 8052 Medical & Biological Engineering & Computing 2000, Vol. 38

conduction to determine the thermal convection (blood flow) component of the probe-tissue heat transfer. For details see the Appendix or VALVANOet al. (1984a).

2.3 Microsphere perfztsion measurement

immediately prior to injection, the labelled microspheres, which are suspended in 20% dextran solution with 0.01% Tween 80 (to act as surfactant against sphere aggregation), were mixed and vortexed thoroughly. The injections each contained approximately 5 x 10 6 spheres § and were given over the course of 30 s through a left atrial catheter. From 10 s before the start of each injection, a reference blood sample was withdrawn at a uniform rate (2mlmin -1) from the femoral artery using a syringe pump** with a glass syringe and the withdrawal continued for 2 min. After animal sacrifice (pentobarbitol overdose while under general anaesthesia), approximately 1 g tissue sample was harvested from the region around the tip of the TDP for activity analysis. The BioPAL microspheres used in this study are in every way like the radioactive microspheres normally used for blood flow studies except that the BioPAL microspheres are made radioactive after tissue harvest. Typical radioactive microspheres, such as those from Dupont NEN, contain isotopes of short halflife rare earth metals imbedded in a capsule. Immediately before these microspheres are shipped from the manufacturer, the short half-life isotopes are created by neutron bombardment of the microspheres. While the BioPAL microspheres also contain rare earth metals imbedded in a capsule, the neutron bombardment, and thus the activation, is performed after tissue harvest when the samples are sent back to BioPAL for counting analysis. Thus, the user never handles any radioactivity and the animal carcass does not contain any radioactivity. In this study, BioPAL stable microspheres labelled with Au, Ir, Sm, Yb and La were used along with Dupont NEN active microspheres labelled with 95Nb and l°3Ru. For analysis, BioPAL irradiates the tissue samples for 10 min at a power level of 1 MW which generates an average neutron flux density of 8 × 1016 s - 1 m - 2 . After activation, samples are stored for 48 h to allow the background radiation of short halflife, artifactual isotopes to decay. Following this decay period, spectrographic analysis is performed to quantify the number of labelled microspheres. This analysis uses a liquid nitrogencooled Ge(Li) detector and corrects for inter-radionuclide crossover and tracer decay during the counting period. The active/ radioactive microspheres were counted using a NaI(TI) gamma well counter*.

The results from the co-injected active and stable microspheres were first used to evaluate the agreement between these two types of spheres. In these animals, an additional biopsy was taken from the liver tissue forcomparative analysis to give a total offourpaired measurements (two injections with two separate tissue samples). Fig. 2 shows a plot of liver blood flow from the stable micro spheres and from the active microspheres. The agreement between the stable and active microspheres is excellent and well within the anticipated repeatability for active/radioactive microspheres. BLAND and ALTMAN (1986) caution against the use of the correlation coefficient (R) alone to determine the agreement between two methods that are hypothesised to measure the same phenomenon. Therefore the results of Fig. 2 are re-plotted in Fig. 3 as a Bland-Altman plot in which the perfusion difference between the two techniques is plotted against the mean perfusion value of the two techniques. This levelo fagreement is generally considered excellent for blood perfusion studies (bias less than 1 mlmin -1 100 g-l). Since the two types ofmicrospheres give statistically equivalent results then the stable/radioactive microspheres may also be considered as valid a 'gold standard' as the unstable/ radioactive microspheres against which to compare the TDP perfusion measurement. While the active/radioactive microspheres provide an acceptable standard for 'snap-shot' blood flow studies in experimental animals, there are several logistical disadvantages. The shorthalf-life isotopes decay over time during storage, these microspheres are costly to purchase, costly to dispose of, and they can potentially be hazardous to laboratory personnel. Fig. 4 shows the validation data for the TDP perfusion measurement in terms of all applicable microsphere data. Liver flow ranged from 0-60 mlmin -1 100 g-1 reflecting animal-toanimal variations. The fit line in Fig. 4 is computed from the leastsquares method (R 2 = 0.898). The slope is very close to tmity (0.921 ) and the intercept is also extremely small (0.566 ml min- 1 100 g-l). Fig. 5 plots the same results as Fig. 4, but in the BlandAltman format with the difference between the microsphere and TDP perfusion shown plotted against the mean value of the microsphere and TDP perfusion. The mean value of the differences, the bias, is very small at 0.623 mlmin -1 100 g-1 Mismatch between the perfusion measured by the microspheres and TDP perfusion is primarily caused by quantisation error, streaming ofmicrospheres (poor mixing), shunt flow and true physiologic perfusion fluctuations. Quantisation error results from the finite number of microspheres in a given 60

"7

oo~ 3 Results and discussion

A total of 27 rabbit experiments were performed with each animal receiving two injections of microspheres, given 30 min apart. Each microsphere injection contained either one active/ radioactive label and one stable/radioactive label ( 12 animals) or two differently labelled stable/radioactive microspheres (15 animals). After animal sacrifice, approximately l g tissue sample was harvested from the region around the tip of the TDP and analysed for paired analysis with the TDP measurement of blood flow. Each microsphere label was considered to give an independent measure of blood flow against which to compare the TDP measurement. §15pm4-0.5pm from BioPAL, Inc., Wellesley, MA and from Dupont New England Nuclear-NEN, North Billerica, MA **Model 94, Harvard Apparatus, Millis, MA *Auto-Gamma 5530, Packard Instruments, Downers Grove, IL

Medical & Biological Engineering & Computing 2000, Vol. 38

"7

40

4l,

.f

E

30

•

• •

•

£ o 20

01

Fig. 2

~

10

~

4'0 ' 20 30 50 -1 -1 activemicrospheres,ml min 100g

' 60

Rabbit liver blood flow as" measured with a co-injection of both active and stable radioactive microspheres to compare results f i r m these two O.'pes o f radiolabels v =

1.0347x 4- 0.031 R2 = 0.964 N = 48 p < 0.0005 321

14

14 "7, "7,

o~

10

o

'7,

'7,.=_

6

l.-

" •~

-2 ~11,*

= "r:

-6

~

-10

-- ~-

mean

•

+2SD=5 2

:

q~'•

• •

• •

.~,

o°~

10

.=_ E

6

0..C3

2

mean + 2 S D = 10.8

y.

mean =-0.47

I---

mean -2SD=-6.1

~ -6

mean =0.623

,

.~_ -2 E

mean -2SD=-9.6

o

.~ 40

•

E -14

I 10

0

I • 20

I 30

I 40

I 50 -1

mean m i c r o s p h e r e perfusion, ml min

I 60

-14

-1

0

•

I

I

I

I

10

20

30

40

I

lOOg

average TDP and micr0sphere, ml min

I

50 -1

60 -1

100g

Fig. 3 Pelfitsion difference as measured with the two micmsphere

Fig. 5 Pelfitsion difference between the TDP and microspheres is

labels is plotted against the mean pelfitsion value for the two labels. The plot is amlotated with the mean difference (Mean) plus and mimts two standard deviations (SD)

plotted against the mean pelfitsion value for the two methods. The plot is mmotated with the mean difference (Mean) plus and mimts two standard deviations (SD)

tissue sample. Given that each label contains 2.5 x 1 0 6 microspheres, that the liver with the portal vein occluded receives approximately 12% of the total cardiac output, and that the liver has a mass of about 40g (an average perfusion of about 30 ml min-1 100 g-l); there should be about 7500 microspheres of the same label in each 1 g of liver tissue. In this case, the activity quantization error or 'shot noise' is 1.2% (7~/7f00/7500). The quantization error is correspondingly higher for lower flow levels. In fact, tissue samples which received no measurable amount ofmicrospheres, corresponding to a flow of less than 0.1mlmin -1 100g -1, were excluded from this analysis because their quamisation error is theoretically infinite (ZwISSLERet al., 1991 ). This was the case for 32 out of all 54 microsphere injections given, it is likely that in this animal model, the clamping of the portal vein produced a shock-like response which then shut down hepatic flow in many animals. Another source of error is the degree to which the microspheres uniformly mix with the blood upon injection, in the protocol, the microspheres were injected directly into the left

atrium to take advantage of the maximal mixing afforded by the blood passing through the atrium and the ventricle. Inadequate mixing would cause the microspheres to preferentially 'stream' into the lobes or part of the liver, out of proportion to the true blood flow. The error induced by possible streaming is difficult to quantify without yet a third independent and simultaneous measure ofperfusion, although its effect is thought to be small. The error due to shunt flow, however, can be estimated by measuring the microspheres in the return venous flow immediately after injection. While we did not do this in our protocol, ZWISSLER et al. (1991) reported that the total systemic shunt flow is 15%, that the myocardial shunt flow is 1%, and that the pulmonary shunt flow is 0.8%. They ascribe the majority of the shunting flow to skin, muscle and fat. We have therefore neglected the effect of hepatic artery shunt flow. VALVANO et al. (1984b) estimated that the microsphere technique in an ex viva rat liver preparation was accurate to within 8% including the effects of streaming, quantisation, radiolabel counting errors and shunting. Comparing the coinjected active and stable microspheres reported in Fig. 3, we estimate the repeatability in the in viva rabbit liver model. Neglecting any errors associated with the radiolabel counting, the root-mean-squared difference between the stable and active microsphere techniques is 2.8mlmin -1 100g -1. Thus for 30 mlmin -1 100 g - l , the repeatability is about 9%. The greatest sources of noise for the TDP are the thermal fluctuations in the liver tissue. Under normal conditions, the animal is able to adequately thermoregulate itself with a mean temperature variation of about 0.025 :C over the course of a measurement sequence. Using the thermal model for perfusion extraction (VALVANO et al., 1984a), this temperature variation causes a perfusion measurement variation of about 10% at a level of 30 mlmin -1 100 g-1. Significant temperature fluctuations in the tissue can also be induced by external manipulations such as injecting cold fluid into the arterial system. The subsequent thermal excursion can cause a perfusion error as the fluid clears through the tissue. To minimise the thermal noise from microsphere injection, TDP measurements were made immediately after the room temperature injectate cleared the liver tissue. Most likely, the greatest part of the disagreement between the TDP perfusion measurements and the microspheres is caused by the fact that the latter provide a 'snap shot' of blood flow. The 'snap shot' is the time it takes to bolus inject the microspheres into the left atrium and subsequently lodge in the vasculature. This time is typically about 5 s. in contrast, the TDP requires an

60

y

50

o

40

E -~ 30

~ 2o 1-

10

0~'~

0

Fig. 4

~

10

I

I

20

30

I

i

60

TDP pelfilsion as a fimction o f the stable and active radioactive microsphere pelfilsion in the rabbit liver model. Data fi'om different microsphere labels in the same co-injection are included in the plot as independent points y = 0 . 9 2 1 x + 0 . 5 6 7 R 2 = 0.898 N = 44 p