micromachined lab-on-a-tube sensors for simultaneous brain temperature and cerebral blood flow...

Micromachined lab-on-a-tube sensors for simultaneous braintemperature and cerebral blood flow measurements

Chunyan Li & Pei-Ming Wu & Jed A. Hartings &

Zhizhen Wu & Cletus Cheyuo & Ping Wang &

David LeDoux & Lori A. Shutter &

Bharat Ram Ramaswamy & Chong H. Ahn &

Raj K. Narayan

Published online: 14 April 2012# Springer Science+Business Media, LLC 2012

Abstract This work describes the development of a micro-machined lab-on-a-tube device for simultaneous measure-ment of brain temperature and regional cerebral blood flow.The device consists of two micromachined gold resistancetemperature detectors with a 4-wire configuration. One isused as a temperature sensor and the other as a flow sensor.The temperature sensor operates with AC excitation current

of 500 μA and updates its outputs at a rate of 5 Hz. The flowsensor employs a periodic heating and cooling techniqueunder constant-temperature mode and updates its outputs ata rate of 0.1 Hz. The temperature sensor is also used tocompensate for temperature changes during the heatingperiod of the flow sensor to improve the accuracy of flowmeasurements. To prevent thermal and electronic crosstalkbetween the sensors, the temperature sensor is located out-side the “thermal influence” region of the flow sensor andthe sensors are separated into two different layers with athin-film Copper shield. We evaluated the sensors for accu-racy, crosstalk and long-term drift in human blood-stainedcerebrospinal fluid. These in vitro experiments showed thatsimultaneous temperature and flow measurements with asingle lab-on-a-tube device are accurate and reliable overthe course of 5 days. It has a resolution of 0.013 °C and0.18 ml/100 g/min; and achieves an accuracy of 0.1 °C and5 ml/100 g/min for temperature and flow sensors respectively.The prototype device and techniques developed hereestablish a foundation for a multi-sensor lab-on-a-tube,enabling versatile multimodality monitoring applications.

Keywords Lab-on-a-tube . Temperature sensor . Cerebralblood flow sensor . Smart catheter . Traumatic brain injury

1 Introduction

Traumatic brain injury (TBI) poses a major public healthproblem and is associated with significant morbidity andmortality. The pathology is heterogeneous, but often leads tosecondary deterioration of the patient’s condition after initialresuscitation and stabilization (Sande and West 2010; Zinket al. 2010). The goal of critical care management is to

C. Li (*) : P.-M. Wu :R. K. NarayanCushing Neuromonitoring Laboratory,Feinstein Institute for Medical Research,North Shore Long Island Jewish Health System,Manhasset, NY 11030, USAe-mail: [email protected]

J. A. Hartings : L. A. ShutterDepartment of Neurosurgery, University of Cincinnati (UC)Neuroscience Institute, UC College of Medicine,Cincinnati, OH 45219, USA

Z. Wu :B. R. Ramaswamy : C. H. AhnMicrosystems and BioMEMS Laboratory, Department ofElectrical and Computer Engineering, University of Cincinnati,Cincinnati, OH 45221, USA

C. Cheyuo : P. WangDepartment of Surgery, Feinstein Institute for Medical Research,North Shore Long Island Jewish Health System,Manhasset, NY 11030, USA

D. LeDouxNeurosurgical Intensive Care Unit,North Shore University Hospital,Manhasset, NY 11030, USA

R. K. NarayanCushing Neuroscience Institute,Hofstra North Shore-LIJ School of Medicine,Hempstead, NY 11549, USA

Biomed Microdevices (2012) 14:759–768DOI 10.1007/s10544-012-9646-7

prevent secondary brain injury by controlling intracranialpressure and maintaining adequate cerebral blood flow andoxygenation (Orliaguet et al. 2008; Hayward and Hunt2011). Advanced monitoring of intracranial variables isessential for the detection of secondary brain injury and toguide medical management. Variables commonly measuredfor clinical care and research include intracranial pressure,jugular venous oxygen saturation, cerebral oxygenation,brain temperature, cerebral perfusion pressure, regionalcerebral blood flow, neurochemistry (through microdialysis)and electroencephalography (EEG) (Stuart et al. 2010;Krajewski et al. 2011; Cecil et al. 2011; Stevens 2004;Hartings et al. 2011; Vidgeon and Strong 2011). Currently,this multimodal monitoring requires intracranial insertion ofmultiple probes and use of several independent monitors,which increases cost, complexity, and patient risk for compli-cations. An ideal monitor would be a single device that cancontinuously provide multiple physiological and biochemicalparameters in real time.

Biomedical microdevices have been used in the clinicalsetting for quite some time and are instrumental to thedelivery of care. Recent developments in microelectrome-chanical systems (MEMS)-based sensors suggest that thequality of diagnosis and treatment can be significantlyenhanced once these advancements are translated into thetargeted applications. The advantages of MEMS-basedsensors over traditional monitoring probes include smallerdimension, ability to integrate, faster response time, lowerpower consumption, lower cost, greater reliability and highersensitivity (Haga and Esashi 2004; Grayson et al. 2004).In addition, the advancement in fabricating the MEMSsensors on a flexible substrate offers a viable solution toone of key technical challenges of rigid monitoringprobes: the mechanical mismatch between the compliantbrain tissue and the sensor substrate (Seymour et al. 2011;Wester et al. 2009).

We have previously developed a novel lab-on-a-tube(LOT), or smart catheter, which is capable of continuouslymonitoring multiple physiological and metabolic parameters(Li et al. 2008, 2009, 2010). Microsensors, wires andcircuits were fabricated first on the flexible polymersubstrate using standard MEMS technology, and thenrolled spirally to make a tube structure. We combinedthe advantages of flexible MEMS technology with aspiral rolling technique to develop a multimodality probeapplicable to monitoring TBI patients.

Monitoring brain temperature (Thompson et al. 2003;Elf et al. 2008) and cerebral blood flow (Dagal and Lam 2011;Mette et al. 2011) play an important role in understanding ofthe pathophysiology of acute brain injury. Fever is common inpatients with TBI, is associated with poor recovery (Jones etal. 1994), and may trigger other mechanisms of secondaryinjury (Hartings et al. 2009). CBF is also a critical variable,

but is usually assessed through the surrogate measure ofcerebral perfusion pressure (mean arterial pressure minusintracranial pressure). This paper describes developmentof a micromachined LOT device for simultaneous braintemperature and cerebral blood flow measurements. Thedevice consists of a temperature sensor based on thin-filmresistance temperature detector (RTD) (Li et al. 2012a) and aflow sensor based on hot wire anemometry (Li et al. 2011a).The micromachined LOT offers several advantages overexisting temperature and flow monitoring approaches.Thin-film RTD can measure temperature as precise as athermocouple and its simple structure is advantageous forsensor integration. Thin-film RTD exhibits superior specifi-cations over the thermistor in terms of linearity, long-termstability, interchangeability, etc (Lipaták 1995). Comparedwith many techniques that are available to assess CBF, suchas near infrared spectroscopy (NIRS), laser doppler flowmetry(LDF), computed tomography (CT) and positron emissiontomography (PET) (Keller et al. 2003; Lam et al. 1997;Schramm 2007; Wintermark et al. 2005), the thermaldiffusion flowmetry-based measurement technique is theonly way that can monitor tissue perfusion continuously,in real time and in absolute physiological units of ml/100 g/min (Rosenthal et al. 2011). The mechanical de-sign and electrical operation of the MEMS-based thinfilm microsensors were carefully chosen such that poten-tial crosstalk resulting from sensor integration becomesnegligible. In addition, we conducted testing in vitro andin human cerebrospinal fluid to evaluate performance andpossible biofouling effects.

2 Design and fabrication

2.1 Materials

Parylene C has superior resistance to moisture absorption(0.06 %) and is qualified as a USP Class VI polymer, thehighest certification level of biocompatibility (Hassler et al.2011). By comparison, Polyimide (Rubehn and Stieglitz2010), another popular polymer for neural implants, hasmuch higher moisture absorption (0.8–1.4 %). However,Polyimide has a much higher tensile modulus (8830 MPa)than Parylene C (20 MPa) and is more robust for sensordevelopment. Based on these properties, we used Polyimideas a support layer and Parylene C as a coating layer todevelop a robust device with superior moisture barrier andbiocompatibility properties.

2.2 Design and working principle

The device consists of two micromachined gold resistancetemperature detectors (RTD) with 4-wire configuration. One

760 Biomed Microdevices (2012) 14:759–768

is used as a temperature sensor (Michalski et al. 2001; Li etal. 2012a) and the other is used as a flow sensor based onhot wire anemometry (Mohamed 2001; Li et al. 2011a). Toprevent possible thermal and electronic crosstalk betweenthe sensors, the temperature sensor is placed outside the“thermal influence” region of the flow sensor and separatedinto two different layers with a thin-film Copper (Cu) shield.Figure 1 shows the design and working principle of thetemperature and flow sensors.

The temperature sensor operates with AC excitation cur-rent of 500 μA and updates its outputs every 200 ms. Itsoutput voltage is derived as follows:

V0 ¼ Vr � 1þ a � ΔTt1ð Þ þ Vn1½ �� Vr � 1þ a � ΔTt2ð Þ þ Vn2½ �

where V0 is the final voltage output, α is the temperaturecoefficient (TCR) of a RTD, Vr is the voltage output at areference temperature (°C), ΔTt1 and ΔTt2 is the deviation intemperature from the reference temperature at time t1 andt2, and ΔVn1 and ΔVn2 represent the voltage offset causedby the thermal electric effect. For fast excitation, it isreasonable to assume that ΔTt10ΔTt20ΔT and ΔVn10ΔVn2.Thus, the equation can be further reduced to:

V0 ¼ 2Vr � 1þ a � ΔTð Þ

Define ΔT0T - Tr, where T is the measured temperatureand Tr is the reference temperature. Rearranging the equationgives the measured temperature:

T ¼ Tr þV02Vr

� 1

a

The flow sensor employs a periodic heating and coolingtechnique under a constant-temperature mode and updatesits outputs every 10 s. During the cooling period, the flowsensor is cooled down to the medium baseline temperatureand during the heating period it is held 2.5 °C above themedium temperature. The detailed operation can be catego-rized into five regions as shown in Fig. 1. The coolingperiod consists of regions I and II and the heating periodcovers regions III, IV, and V. In region I, no current isapplied and the sensor is cooled down from previous heatingperiod. In region II, after the sensor settles to the mediumtemperature, a small current (500 μA) is applied, the sensorresistance is measured, and the overheating resistance iscalculated. In region III, the circuit initiates the feedbackcontrol, thereby inducing an overshoot during the course ofheating the sensor. Meanwhile, the peak output is sampledfor use in the thermal conductivity compensation (Li et al.2011a, 2012b). The output then settles down to a stablevalue in region IV. During the heating period, mediumtemperature compensation is achieved by adjusting the

Flow sensor

Temperature sensor

I. Fully cool down (zero current) II. Baseline temperature measurement III. Medium thermal conductivity measurement IV. 2.5ºC increment with temperature compensation V. Sampling with thermal conductivity compensation

Cooling period

Heating period

I

III

IV

V

Pseudo steady-state

Steady-stateII

150µm

25µm

900µm

775µm

25µm

100µm

575µm

5000µm300µm 1500µm

I (excitation current)

+500µA

-500µA

Timet1

t2

100ms

V0 = (R(t1) – R(t2)) ·I

<Cross Section View>

7.5µm Kapton 1µm Parylene

Ti/Au 150/1200Åtemperature sensor

Ti/Au150/1200Åflow sensor

2µm Parylene

2µm Parylene

3µm Parylene

1500Å Cu

Fig. 1 Micromachined lab-on-a-tube (LOT) with brain temperaturesensor and cerebral blood flow sensor: Temperature sensor with 4-wireresistance temperature detector (RTD) configuration operates with acexcitation current of 500 μA and updates its outputs every 200 ms.

Periodic cooling and heating technique is used for the flow sensor to doin situ temperature and thermal conductivity compensation. It operatesin a constant-temperature mode and updates its outputs every 10s

Biomed Microdevices (2012) 14:759–768 761

overheating resistance according to the temperature sensoroutput. The end of region IV when steady state is achievedis defined as region V. Here, multiple data points are sam-pled and their average represents the flow rate withoutthermal conductivity compensation. The output withoutthermal conductivity compensation correlates to the flowrate using the simplified equation (Chen and Liu 2003):

Vf ¼ C þ D � Fn

where Vf is the uncompensated voltage output, C and D arecoefficients related to conduction and convection, F is theflow rate and n is the fitting factor. The flow rate will befurther processed with medium thermal conductivity com-pensation using the information from region III.

2.3 Fabrication

Polydimethylsiloxane (PDMS, Sylgard 184A and 184B,DOW corning) was spin-coated on a six inch Silicon waferand cured at 70 °C for 1 h. A 7.5 μm thick Kapton film (typeHN) was cleaned and attached to the PDMS surface. 1 μmthick Parylene film was deposited (Model PDS 2010 Lab-coter 2, Specialty Coatings Systems, USA) to smooth thefilm surface and also enhance the resistance to moisturetransmission (Menon et al. 2009; Hassler et al. 2010; Hsuet al. 2009). Electrodes for the temperature sensor werefabricated by depositing and patterning a Au (1,200 Å) layerwith an adhesion layer of Ti (150 Å) using an E-beam

evaporator (Temescal FC1800, BOC Edwards Temescal,USA), followed by standard lithography and etching processes.After that, the electrical leads of the temperature sensor wereelectroplated with 2 μm thick Cu to reduce the lead resistances.Then 2 μm Parylene, 1,500 Å Cu, and 2 μm Parylene weredeposited in sequence to prepare the surface to develop the flowsensor. The electrodes for the flow sensor were fabricated bydepositing and patterning the 150 Å Ti /1200 Å Au layers. Theelectrical leads of the flow sensor were also electroplated with2 μm thick Cu to reduce the lead resistances. Finally, 3 μmthick Parylene film was deposited as an outermost layer.

The film with the temperature and flow sensors were cutinto size (width02.5 mm; length0150 mm) and spirallyrolled over the metal rode (ASIN B00137YCEE, SmallParts Inc., USA) based on our previous work to form anintraventricular catheter (Li et al. 2009). After spiral-rolling,a post-treatment (electrified with a 20 mA current for 12 h)was performed in order to enhance stability and also alleviatestress induced during the rolling procedure. A fabricated LOTdevice with the temperature and flow sensors is shown inFig. 2.

2.4 Measurement system

Figure 2 illustrates the deployment of the LOT temperatureand flow sensors in practice. The LOT device is connectedto the interface circuits through a 1 m long flexible cable sothat the interface circuit can be installed at the bedside. Theinterface circuit consists of two printed circuit boards (PCB)

Multimodality Monitor: LabView Program

in the Laptop Computer

Power Supply and DAQ

Interface Circuits

Micromachined Lab-on-a-Tube for Multimodality monitoring

Enlarged View for Temperature and Flow Sensors on the Tube

Temperature sensor Flow sensor

Cu electroplated leads for flow sensor

Connector

Cu electroplated leads for temperature sensor

Sensing region with temperature and flow sensors

DAQ

Power supply

PCB for temperature sensor PCB for flow sensor

Connection through USB cable

Connection through 3 m long cable

Connection through 1m long

cable

Fig. 2 Developed LOT prototype and its bedside temperature andflow monitoring system: the micromachined LOT device with micro-sensors and the container with interface circuits are connected with 1 mlong flexible cable. The container with interface circuits is then

connected to the container which contains the power supply andDAQ through 3 m long cables. The DAQ is then connected to themultimodality monitor which is programmed with LabView softwarethrough the USB cable. The whole system is portable


for respective sensors. The circuits are placed inside a metalenclosure (H: 10.5 cm, W: 13.4 cm, L: 22.1 cm) to mitigateelectromagnetic interference (EMI). The associated powersupply is located in another enclosure (H: 7.6 cm, W:18.2 cm, L: 28.4 cm) along with the data acquisition(DAQ) device. Interconnection between the circuit and thepower supply is 3 m long. The multimodality monitor (Li etal. 2012a) connected to the DAQ via a USB cable sitsdirectly on the top of the box. The whole system is designedto be portable.

The temperature sensor circuit performs the functions ofcurrent excitation, voltage readout, analog-to-digital (A/D)conversion, signal conditioning and calibration. The sensoris driven by a square wave current with constant amplitudeof 500 uA. The AC output voltage of the sensor is capturedby the A/D converter and demodulated digitally using themicrocontroller (MCU). The final output in units of temper-ature is generated from the MCU. The flow sensor circuitshares similar architecture to the temperature sensor circuitexcept that the current source is variable. The current re-quired to maintain the feedback serves as the basic indicatorfor the flow rate. The current is converted into the compen-sated output in units of absolute flow rate via the MCU.

The front-end circuit of each sensor has its own isolatedground. This aims to prevent unwanted ground noise or loopcurrent from degrading the sensitive front-end circuit. Iso-lation between the front-end and back-end circuits is real-ized by digital isolators. All the back-end circuits and theDAQ share the same ground. The ground noise at the backend is not a significant concern since the output signals atthis point have already been digitized or amplified.

3 Experimental results

A total of ten LOT devices were evaluated. Before themeasurements, LOT devices were sterilized with ethyleneoxide gas (EtO) at the North Shore University Hospital.They were then connected to the developed multimodalitymeasurement system for data collection. Blood-stained ce-rebrospinal fluid (CSF) drained from TBI patients in thecourse of their treatment was used to assess potential effectof biofouling on the sensitive surface of the temperature andflow sensors, which could influence performance. A tissueperfusion simulator (Thalayasingam and Delpy 1989) wasconstructed to measure the outputs under different flow rates(Li et al. 2011a). In the simulator, water flows from aconstant pressure head reservoir through a metal tube andinto the perfusion chamber containing a piece of sponge(10 ml) so that flow through the sponge is radially outwards.The sponge is used to simulate tissue and the metal tubeis used to provide a uniform temperature at the waterinlet. A precision valve (Model:VM3-TT-0AA, AALBORG,

USA) is used to precisely control the flow rate and the wholesystem is housed in a cell culture incubator (Model: 5410-L,NAPCO, USA) to maintain temperature in the physiologicrange.

3.1 Timing diagram

The LOT temperature sensor operates with AC excitationcurrent of 500 μA. Its output voltage as a result of the ACexcitation is shown in Fig. 3(a), where its peak-to-peakvalue correlates linearly to the temperature. The LOT tem-perature sensor updates the outputs every 200 milliseconds.

When the temperature of the electrically heated LOTflow sensor increases, it loses power to its surrounding untilit reaches a thermal equilibrium. An optimal compromise ofhigh temporal resolution and output accuracy is achieved bya cycle of 6 s cooling and 4 s heating (Li et al. 2011a).Figure 3(b) shows the sensor output during a complete

Time (seconds)

0

100

200

300

400

500

600

700

0 1 2 3 4 5 6 7 8 9 10

Vol

tage

out

puts

(m

V)

Time (seconds)

I

II

III

IVV

Cooling Period

Heating Period

Vol

tage

out

puts

(m

V)

0 1 2 3 4 5 6 7 8 9 10

0

475

950

1900

-475

-950

-1900

(a)

(b)

Fig. 3 Timing diagram: (a) Temperature sensor operates with acexcitation current of 500 μA and 5 Hz. The outputs update every200 ms. (b) Flow sensor operates with periodic cooling and heatingtechnique. The cooling period (I, II) is 6 s and heating period (III, IV, V) is4 s. The outputs update every 10s


cooling-heating cycle with the five operation regions. Thedetailed function of each region is given in the design andworking principle section.

3.2 Crosstalk between temperature and flow sensors

In the integrated system, thermal and electrical crosstalkbetween the sensors can be a significant problem and mustbe reduced below a certain admissible level. To preventthermal crosstalk that originates from heating the flowsensor, the temperature sensor is located outside the“thermal influence” region of the flow sensor at a distanceof 1.5 mm. Thermal profile of the flow sensor was simulatedusing ANSYS v13. Figure 4(a) gives the 2D plot in x-y plane.The flow sensor exhibits a uniform temperature of 2.5 °Cabove the ambient temperature, 36 °C. The result is acquiredafter the heat is applied for 4 s, indicating a thermal influencerange 1 mm. The distance is also advantageous in keeping thecapacitive and inductive coupling low. To further protect thesensors from the electromagnetic interference, they are ar-ranged into two different layers and separated with a 2 μmParylene/1500 Å Cu/2 μm Parylene barrier. The Cu film canact as an EMI shield.

Figure 4(b) shows the temperature sensor outputs withthe flow sensor switched on and off alternately. The flowsensor is turned on at 60, 180, and 300 s and turned off at120, 240, and 360 s. Under the standard design (1.5 mmdistance), the result indicates that no significant crosstalkoccurs between the two sensors since no abrupt changes areobserved during the switching events. We repeated thesemeasurements under a different configuration (proximity)where the two individual sensors are placed face-to-facewith a distance of 1 mm. Thermal influence becomes appre-ciable under this condition. Figure 4(c) shows the flowsensor outputs as the temperature sensor is switched onand off alternately with the same timing and conditions asin the previous test. Regardless of the conditions, the flowsensor’s output fluctuation, as much as 2.5 ml/100 g/min, isstill within the regular noise level. The results confirm thatunder the standard design the flow sensor operation is notaffected by the temperature sensor’s status.

3.3 Temperature sensor: flow rate effect

For high accuracy temperature measurement, self-heatingerrors should always be taken into account. In our previouswork (Li et al. 2012a), we optimized the excitation currentvalue to prevent self-heating of the sensor, but also achieveadequate signal-to-noise ratio (SNR). We tested the temper-ature sensors under the flow system ranging from 0 to160 ml/100 g/min. If the temperature sensor has a self-heating effect, its outputs should decrease when the flowrates increase.

Figure 5 shows the experimental results. The maximumerror is smaller than 0.03 °C, which adds little error to the

Time (seconds)

Flow

rat

e (m

l/100

g/m

in)

(c)

(a)

1mm

ON

OFF

ON

ON

OFF

OFF

49.5

50

50.5

51

51.5

52

52.5

53

0 60 120 180 240 300 360 420 480

Proximity

Standar

(b)

Time (seconds)

Tem

pera

ture

(°C

)

ON

OFF

ON

OFF ON

OFF

36.76

36.78

36.8

36.82

36.84

36.86

36.88

36.9

36.92

0 60 120 180 240 300 360 420 480

Proximity

Standard

Fig. 4 Crosstalk evaluation between temperature and flow sensors: (a)ANSYS thermal plot of the flow sensor after heating 2.5 °C above theambient temperature for 4 s. (b) Temperature was measured with theflow sensor switched on and off alternately every 60 s. (c) Flow wasmeasured with the temperature sensor switched on and off alternatelyevery 60 s. In standard case, the two sensors are positioned followingthe design in Fig. 1, whereas in proximity case, two distinct sensors arepositioned face-to-face with a distance of 1 mm. The temperaturesensor output is affected by the thermal influence of the flow sensorin proximity case. In standard case, there is no significant crosstalk ateither sensor’s output due to the switching behavior


desired temperature sensor accuracy of 0.1 °C. Thus, oper-ation of the temperature sensor with 500 μA ac excitationcurrent does not cause any self-heating in the flow range from0 to 160 ml/100 g/min.

3.4 Flow sensor: overheating temperature

The flow sensor was calibrated under different flow rates(from 0 to 100 ml/100 mg/min) at three distinct overheatingtemperatures (2 °C, 2.5 °C, 3 °C). Figure 6 shows the flowsensor sensitivity at three different overheating temperatures.The experimental results show that flow sensor sensitivityincreases as the overheating temperature increases with an

increment of ~15–20 % per 0.5 °C increase. The commercialthermal diffusion probe (Qflow500, Hemedex Inc.) for cere-bral blood flow measurements uses an overheating tempera-ture of 2.5 °C (Vidgeon and Strong 2011), so we chose 2.5 °Cas the overheating temperature for the LOT flow sensor. TheLOT flow sensor has a sensitivity of 78.56±3.9 mV/ml/100 g/min in the range of 0 to 160 mV/ml/100 g/min.

3.5 Flow sensor: temperature compensation

The periodic heating operation method itself can compen-sate for changes in medium temperature that occur duringmonitoring (i.e. through successive periods of cooling-heating) (Li et al. 2011b). To compensate for temperaturechanges occurring during the heating period (region IV), amethod utilizing the temperature sensor was investigated (Liet al. 2011a). The temperature sensor is used to sense thetemperature variation during the heating period and thereadings were used to adjust the overheating resistance byusing the following equation:

Ra ¼ Ri � 1þ Th � TiTi

� Gi

� �

where Ra is the adjusted overheating resistance, Ri the initialoverheating resistance, Th the temperature measured duringthe heating period, Ti the initial temperature of the heatingperiod, andGi the gain specific for the temperature compensation.

Figure 7 compares the compensated and uncompensatedflow sensor outputs plotted against the medium temper-ature change. The sensor was subjected to the flow rateof ~50 ml/100 g/min and the medium temperature decreased

Flow rate (ml/100g/min)

Res

ista

nce

(Ohm

)

Tem

pera

ture

mea

sure

d (°

C)

102.46

102.465

102.47

102.475

102.48

102.485

102.49

102.495

102.5

102.505

0 20 40 60 80 100 120 140 160 18036.465

36.47

36.475

36.48

36.485

36.49

Fig. 5 The effect of flow on the temperature measurement: Thetemperature sensor outputs were recorded under different flow rates.The outputs are almost independent of the flow rates ranging from 0 to160 ml/100 g/min. The maximum error is smaller than 0.03 °C

Overheating temperature (°C)

Sens

itivi

ty (

mV

/ml/1

00g/

min

)

50

55

60

65

70

75

80

85

90

95

100

1.5 2 2.5 3 3.5

Fig. 6 Overheating temperature versus flow sensor sensitivity: Theflow sensor sensitivity was measured under the overheating temperatureof 2 °C, 2.5 °C and 3 °Cwithin the flow rates from 0 to 100ml/100 g/min.The outputs sensitivity increases with the overheating temperatureincrease

Time (seconds)

Flow

rat

e (m

l/100

g/m

in)

Tem

pera

ture

mea

sure

d (°

C)

40

42

44

46

48

50

52

54

56

58

60

0 200 400 600 800 1000 30.5

31

31.5

32

32.5

33

33.5

34

34.5

35

35.5 With compensation

Without compensation

Temperature

Fig. 7 Temperature compensation of flow sensor during the heatingperiod: The flow sensor was subject to the flow of ~50 ml/100 g/minwhereas the medium temperature decreased at a steady rate of ~0.025 °C per 5 s. The uncompensated outputs on average were 1.5 ml/100 g/min higher than the compensated outputs


at a steady rate of ~0.025 °C per 5 s. On average the uncom-pensated outputs were 1.5 ml/100 g/min higher than thecompensated outputs. The results agree with the theory thatflow output increases if overheating resistance is fixed andmedium temperature decreases during the heating period.Temperature compensation can thus reduce the error causedby medium temperature variation during the heating period.

3.6 Flow sensor: thermal conductivity

It has been found that the medium thermal conductivity iswell correlated to the squared peak current during heatingperiod (region III) (Li et al. 2011a, 2012b). During opera-tion, three data points associated with the peak outputs weresampled and their average was stored to account for thecontribution of medium thermal conductivity to the firstorder. The compensated flow rate, Fc, is generated fromthe following linear equation:

Fc ¼ Vu � Vp2 � Gc � V0

S

where Vu is the uncompensated output, Vp is the averagepeak output, Vc is the gain specific for the thermal conduc-tivity compensation, Vo is the baseline voltage, and S thesensor sensitivity. Gc and Vo are derived experimentallyunder various medium thermal conductivities, and underno flow condition, respectively.

Figure 8 shows the uncompensated and compensatedflow sensor outputs versus medium thermal conductivity.The flow sensor was under a constant flow rate of ~50 ml/

100 g/min. Glucose solutions of different concentrationswere used to produce media with different thermal conduc-tivities. The results indicate that using the squared peakcurrents to compensate the flow sensor outputs can reducethe error from 8.5 ml/100 g/min to 1 ml/100 g/min when themedium undergoes thermal conductivity changes of0.17 W·m−1·K−1.

3.7 Long-term drift

LOT temperature and flow sensors were tested in theclinical blood-stained CSF solution to assess time-dependentdrift in temperature and flow rate recordings. Data were

Thermal conductivity (W·m-1·K-1)

Flow

rat

e (m

l/100

g/m

in)

47

49

51

53

55

57

59

61

63

0.4 0.42 0.44 0.46 0.48 0.5 0.52 0.54 0.56 0.58 0.6

Without compensation

With compensation

Fig. 8 Thermal conductivity compensation during the heating period:The flow sensor is subject to the constant flow rate of ~50 ml/100 g/minand the solution with different thermal conductivity was introduced.Using squared peak currents to compensate the flow sensor outputs canreduce the error from 8.5 ml/100 g/min to 1 ml/100 g/min when themedium thermal conductivity was changed from 0.59 W·m−1·K−1 to0.42 W·m−1·K−1

LOT temperature sensor

Commercial temperature sensor

Com

mer

cial

tem

pera

ture

sen

sor

(°C

)

LO

T te

mpe

ratu

re s

enso

r (°

C)

37.6

37.4

37.2

37

36.8

36.6

36.4

36.2

36

37.6

37.4

37.2

37

36.8

36.6

36.4

36.2

36

Time (hours)

(a)


Commercial temperature sensor

0 20 40 60 80 100 120

(b) 37.6

37.4

37.2

37

36.8

36.6

36.4

36.2

36

Time (hours)

LO

T te

mpe

ratu

re s

enso

r (°

C)

LO

T f

low

sen

sor

(ml/1

00g/

min

)

58

52

50

48

46

56

54

60

0 20 40 60 80 100 120


LOT flow sensor


LOT flow sensor

Fig. 9 Long-term drift tests in the clinical CSF solution: The accuracyof the temperature and flow sensors were measured continuously for5 days in blood-stained CSF solution at the temperature of 36.8±0.5 °Cand flow rate of 50±5 ml/100 g/min. The data for 5 days are shown. (a)The outputs from the LOT temperature sensor and commercialtemperature sensor show that LOT temperature sensor follows wellwith the commercial sensor. (b) The outputs from LOT flow sensorshow that flow sensor keeps its accuracy in the temperature changingenvironment


recorded continuously for 5 days. A commercial temperatureprobe, IT-21 (WPI, USA), which is a T type clinicalmicroprobe with guaranteed 0.1 °C accuracy, was inserted tosimultaneously record the medium temperature changes.

Figure 9(a) shows the experimental data continuouslyrecorded for 5 days for the LOT temperature sensor. TheLOT temperature sensor follows well with the commercialIT-21 probe in the temperature changing environment. TheLOT temperature sensor has a resolution of 0.013 °C andachieves the accuracy of 0.1 °C. Figure 9(b) shows theexperimental data continuously recorded for 5 days for theLOT flow sensor. With in situ temperature and thermalconductivity compensation, the LOT flow sensor has aresolution of 0.18 ml/100 g/min and achieves the accuracybetter than 5 ml/100 g/min over the course of 5 days. Theseresults also indicate that if minimal biofouling does occurwith the LOT device as for other implantable probes, it willnot affect the accuracy or reliability of either sensor.

4 Conclusion

In this work, we have presented the design, operation andperformance of a micromachined LOT device with the tem-perature and flow sensors. Thermal and electronic crosstalkbetween the sensors was greatly reduced by locating thetemperature sensor outside the “thermal influence” region ofthe flow sensor and by placing the sensors on two differentlayers with a thin-film Cu shield. Under the working mecha-nism of AC excitation of the temperature sensor and periodicheating and cooling of the flow sensor, simultaneous temper-ature and flow rate measurements with a single LOT devicewere shown to be accurate and reliable in human blood-stained CSF solution over the course of 5 days. Since theLOT microsensors show satisfactory results and have alsopassed the mechanical, electrical and in vitro cytotoxicity testswhich we plan to report in our future publication, they holdpromise for in vivo evaluation and clinical use. Although LOTdevice focuses on intracranial monitoring for acute braininjury, we expect that the techniques and prototype developedhere can be applied to a variety of tissue monitoring purposes.

Acknowledgments This work was supported by the US Departmentof Defense under project No. PT090526P4 (W81XWH-10-1-0978).The authors gratefully acknowledgeMr. Melvin O. Gonzalez at the NorthShore University Hospital for performing polyethylene gas sterilization.

References

S. Cecil, P.M. Chen, S.E. Callaway, S.M. Rowland, D.E. Adler,J.W. Chen, Traumatic brain injury: Advanced multimodalneuromonitoring from theory to clinical practice. Crit. CareNurse 31, 25–37 (2011)

J. Chen, C. Liu, Development and characterization of surfacemicromachined, out-of-plane hot-wire anemometer. J. Micro-electromech. Syst. 12, 979–988 (2003)

A. Dagal, A.M. Lam, Cerebral blood flow and the injured brain howshould we monitor and manipulate it? Curr. Opin. Anaesthesiol.24, 131–137 (2011)

K. Elf, P. Nilsson, E. Ronne-Engström, T. Howells, P. Enblad, Tem-perature disturbances in traumatic brain injury: Relationship tosecondary insults, barbiturate treatment and outcome. Neurol.Res. 30, 1097–1105 (2008)

A.C.R. Grayson, R.S. Shawgo, A.M. Johnson, N.T. Flynn, L. Yawen,M.J. Cima, R. Langer, A BioMEMS review: MEMS technologyfor physiologically integrated devices. P IEEE 92, 6–21 (2004)

Y. Haga, M. Esashi, Biomedical Microsystems for minimally invasivediagnosis and treatment. P IEEE 92, 98–114 (2004)

J.A. Hartings, A.J. Strong, M. Fabricius, A. Manning, R. Bhatia, J.P.Dreier, A.T. Mazzeo, F.C. Tortella, M.R. Bullock, Spreadingdepolarizations and late secondary insults after traumatic braininjury. J. Neurotrauma 26, 1857–1866 (2009)

J.A. Hartings, M.R. Bullock, D.O. Okonkwo, L.S. Murray, G.D. Murray,M. Fabricius, A.I. Maas, J. Woitzik, O. Sakowitz, B. Mathern, B.Roozenbeek, H. Lingsma, J.P. Dreier, A.M. Puccio, L.A. Shutter, C.Pahl, A.J. Strong, Spreading depolarisations and outcome aftertraumatic brain injury: a prospective observational study. LancetNeurol. 10, 1058–1064 (2011)

C. Hassler, R.P. Von Metzen, P. Ruther, T. Stieglitz, Characterization ofparylene C as an encapsulation material for implanted neuralprostheses. J. Biomed. Mater. Res. A 93B, 266–274 (2010)

C. Hassler, T. Boretius, T. Stieglitz, Polymers for neural implants. J.Polymer. Sci. Polymer. Phys. 49, 18–33 (2011)

E. Hayward, K. Hunt, Clinical neuroprotection and secondary neuronalinjury mechanisms. Anaesth. Intensive Care Med. 12, 198–200(2011)

J.M. Hsu, L. Rieth, R.A. Normann, P. Tathireddy, F. Solzbacher,Encapsulation of an integrated neural interface device withParylene C. IEEE Trans. Bio-Med. Eng. 56, 23–29 (2009)

P.A. Jones, P.J.D. Andrews, S. Midgley, S.I. Anderson, I.R. Piper, J.L.Tocher, A.M. Housley, J.A. Corrie, J. Slattery, N.M. Dearden,J.D. Miller, Measuring the burden of secondary insults in head-injured patients during intensive care. J. Neurosurg. Anesthesiol.6, 4–14 (1994)

E. Keller, A. Nadler, H. Alkhadi, S. Kollias, Y. Yonekawa, P.Niederer, Non invasive measurement of regional cerebral bloodflow and regional cerebral blood volume by near infrared spec-troscopy and indocyanine green dye dilution. Neuroimage 20,828–839 (2003)

K.L. Krajewski, B. Orakcioglu, D. Haux, D.N. Hertle, E. Santos, K.L.Kiening, A.W. Unterberg, O.W. Sakowitz, Cerebral microdialysisin acutely brain-injured patients with spreading depolarizations.Acta Neurochir. Suppl. 110, 125–130 (2011)

J.M. Lam, J.N. Hsiang, W.S. Poon, Monitoring of autoregulation usinglaser Doppler flowmetry in patients with head injury. J. Neuro-surg. 86, 438–445 (1997)

C. Li, P.M. Wu, J. Han, C.H. Ahn, A flexible polymer tube lab-chipintegrated with microsensors for smart microcatheter. Biomed.Microdevices 10, 671–679 (2008)

C. Li, P.M. Wu, W. Jung, C.H. Ahn, L.A. Shutter, R.K. Narayan, Anovel lab-on-a-tube for multimodality neuromonitoring of patientswith traumatic brain injury (TBI). Lab Chip 9, 1988–1990 (2009)

C. Li, L.A. Shutter, P.M. Wu, C.H. Ahn, R.K. Narayan, Potential of asimple lab-on-a-tube for point-of-care measurements of multipleanalytes. Lab Chip 10, 1476–1479 (2010)

C. Li, P.M. Wu, J.A. Hartings, Z. Wu, C.H. Ahn, D. LeDoux, L.A.Shutter, R.K. Narayan, Smart catheter flow sensor for real-timecontinuous regional cerebral blood flow monitoring. Appl. Phys.Lett. 99, 233705-1-233705-4 (2011a)


Li, P.M. Wu, Z. Wu, C.H. Ahn, J.A. Hartings, R.K. Narayan, SmartCatheter Flow Sensor for Continuous Regional Cerebral BloodFlow Monitoring. Proc. IEEE Conference on Sensors (Limerick,Ireland, October 28–31, 2011b), pp. 1417–1420

C. Li, P.M. Wu, Z. Wu, C.H. Ahn, D. LeDoux, L.A. Shutter, J.A.Hartings, R.K. Narayan, Brain temperature measurement: A studyof in vitro accuracy and stability of smart catheter temperaturesensors. Biomed. Microdevices 14, 109–118 (2012a)

C. Li, P.M. Wu, J.A. Hartings, Z. Wu, C.H. Ahn, R.K. Narayan,Cerebral Blood Flow Sensor with in Situ Temperature and ThermalConductivity Compensation. Proc. IEEE MEMS Conference(Paris, France, January 29-February 2, 2012b), pp. 1021–1024

B.G. Lipták, Process measurement and analysis. CRC Press, 690(1995)

P.R. Menon, W. Li, A. Tooker, Y.C. Tai, Characterization of watervapor permeation through thin film Parylene C. Proc of the 15thinternational conference on Solid-State Sensors, Actuators andMicrosystems, 2009, pp. 1892–1895

D. Mette, R. Strunk, M. Zuccarello, Cerebral blood flow measurementin neurosurgery. Trans. Stroke Res. 2, 152–158 (2011)

Michalski, K. Eckersdorf, J. Kucharski, J. Mcghee, Temperature Mea-surement. (John Wiley & Sons Ltd, 2001), pp. 88–90

G. Mohamed, The MEMS Handbook. (CRC Press, 2001), pp. 2615–2617

G.A. Orliaguet, P.G. Meyer, T. Baugnon, Management of critically illchildren with traumatic brain injury. Paediatr. Anaesth. 18, 455–461 (2008)

G. Rosenthal, R.O. Sanchez-Mejia, N. Phan, J.C. Hemphill, C. Martin,G.T. Manley, Incorporating a parenchymal thermal diffusion ce-rebral blood flow probe in bedside assessment of cerebral autor-egulation and vasoreactivity in patients with severe traumaticbrain injury. J. Neurosurg. 114, 62–70 (2011)

B. Rubehn, T. Stieglitz, In vitro evaluation of the long-term stability ofpolyimide as a material for neural implants. Biomaterials 31,3449–3458 (2010)

A. Sande, C. West, Traumatic brain injury: a review of pathophysiologyand management. J. Vet. Emerg. Crit. Care 20, 177–190 (2010)

P. Schramm, High-concentration contrast media in neurologicalmultidetector-row CTapplications: implications for improved patientmanagement in neurology and neurosurgery. Neuroradiology 49,S35–S46 (2007)

J.P. Seymour, N.B. Langhals, D.J. Anderson, D.R. Kipke, Novelmulti-sided, microelectrode arrays for implantable neuralapplications. Biomed. Microdevices 13, 441–451 (2011)

W.J. Stevens, Multimodal monitoring: Head injury management usingSjvO2 and LICOX. J. Neurosci. Nurs. 36, 332–339 (2004)

R.M. Stuart, M. Schmidt, P. Kurtz, A. Waziri, R. Helbok, S.A. Mayer,K. Lee, N. Badjatia, L.J. Hirsch, E.S. Connolly, J. Claassen,Intracranial multimodal monitoring for acute brain injury: asingle institution review of current practices. Neurocrit. Care12, 188–198 (2010)

S. Thalayasingam, D.T. Delpy, Thermal clearance blood flowsensor-sensitivity, linearity and flow depth discrimination.Med. Biol. Eng. Comput. 27, 394–398 (1989)

H.J. Thompson, N.C. Tkacs, K.E. Saatman, R. Raghupathi, T.K.McIntosh, Hyperthermia following traumatic brain injury: acritical evaluation. Neurobiol. Dis. 12, 163–173 (2003)

S.D. Vidgeon, A.J. Strong, Multimodal cerebral monitoring intraumatic brain injury. J. Iran Chem. Soc. 12, 126–133 (2011)

B. A. Wester, R. H. Lee, M. C. Laplaca, Development and character-ization of in vivo flexible electrodes compatible with large tissuedisplacements. J. Neural. Eng. 6, 024002-1-024002-7 (2009)

M. Wintermark, M. Sesay, E. Barbier, K. Borbély, W.P. Dillon, J.D.Eastwood, T.C. Glenn, C.B. Grandin, S. Pedraza, J.F. Soustiel, T.Nariai, G. Zaharchuk, J.M. Caillé, V. Dousset, H. Yonas,Comparative overview of brain perfusion imaging techniques.Stroke 36, e83–e99 (2005)

B.J. Zink, J. Szmydynger-Chodobska, A. Chodobski, Emergingconcepts in the pathophysiology of traumatic brain injury.Psychiatr. Clin. North Am. 33, 741–756 (2010)


micromachined lab-on-a-tube sensors for simultaneous brain temperature and cerebral blood flow...

Documents