Design of an fMRI-Compatible Optical Touch Stripe Based onFrustrated Total Internal Reflection

Behnaz Jarrahi1,2, Student Member, IEEE, and Johann Wanek3

AbstractPreviously we developed a low-cost, multi-configurable handheld response system, using a reflective-typeintensity modulated fiber-optic sensor (FOS) [1] to accuratelygather participants behavioral responses during functionalmagnetic resonance imaging (fMRI). Inspired by the popularityand omnipresence of the fingertip-based touch sensing userinterface devices, in this paper we present the design of aprototype fMRI-compatible optical touch stripe (OTS) as analternative configuration. The prototype device takes advantageof a proven frustrated total internal reflection (FTIR) technique.By using a custom-built wedge-shaped optically transparentacrylic prism as an optical waveguide, and a plano-concavelens to provide the required light beam profile, the positionof a fingertip touching the surface of the wedge prism canbe determined from the deflected light beams that becometrapped within the prism by total internal reflection. Toachieve maximum sensitivity, the optical design of the wedgeprism and lens were optimized through a series of lightbeam simulations using WinLens 3D Basic software suite.Furthermore, OTS performance and MRI-compatibility wereassessed on a 3.0 Tesla MRI scanner running echo planarimaging (EPI) sequences. The results show that the OTS candetect a touch signal at high spatial resolution (about 0.5 cm),and is well suited for use within the MRI environment withaverage time-variant signal-to-noise ratio (tSNR) loss< 3%.

I. INTRODUCTION

With an increasing popularity of magnetic resonanceimaging (MRI), efficient and high performing devices thatare compatible with use in an MRI setting are becomingmore in demand. In recent years, interfaces with user-friendlyand intuitive navigation options such as touch sensing foundin small display devices (e.g. Apples iPhone [2], and iPodtouch [3]) and tabletop systems (e.g. Microsoft Surface[4], SMART Table [5], etc.) have become commonplace ineveryday life. Several touch sensing mechanisms have beendeveloped including resistive membrane [6], surface acousticwave [7], piezoelectric [8], capacitive [9] and optical [10],[11] sensing techniques. However, utilizing such interfaces inMRI is still a challenge. The strong magnetic environment ofthe scanner limits the use of the conventional electromagneticequipments. In addition, the current-carrying conductors areparticularly liable to produce image artifacts by distortingthe magnetic field homogeneity, especially when they are inthe close proximity of the magnet [12][14].

1,2B. Jarrahi is with the Department of Information Technology andElectrical Engineering, Swiss Federal Institute of Technology (ETH) Zurich,CH-8049 Zurich, Switzerland, and also with the Institute of Neuroradiology,University Hospital Zurich, CH-8091 Zurich, Switzerland; Phone: +41 44255 56 03; (email: jarrahib@ethz.ch).

3J. Wanek is with the Spinal Cord Injury Center, BalgristUniversity Hospital, CH-8008 Zurich, Switzerland; (email:jwanek@paralab.balgrist.ch).

Among the available touch sensing approaches, opticaltouch sensing is more promising for implementation inMRI especially due to its inherent invulnerability toelectromagnetic interference. But conventional optical touchsystems primarily use infrared (IR) vision-based sensingwhich requires an IR imaging camera to be positioned in rearor in front of the touch surface for the direct line-of-sight tothe objects being sensed [15]. Considering the characteristicsof MRI environment, compatibility of electro-optical sensorshave to be verified to avoid disturbance to the MRI system.Although there are a few existing MRI-compatible touch

pad systems (e.g. [16]), they are generally costly or stilluse electrical cables to communicate with the equipmentsoutside the MRI scanner room. Accordingly, in this paper thedesign of a low-cost optical touch stripe (OTS) is presented.The sensing unit (Fig 1, top), which is completely freeof electrical wires and ferromagnetic materials, is basedon optical detection of internally reflected light beams.Preliminary results show that it is MRI compatible and atechnically feasible user interface for applications in fMRI.

Fig. 1. The OTS prototype (a) with the light beams from a fiber-optic sensorilluminating the inside of the wedge-shaped acrylic prism in ambient light,(b) in the dark, and (c) with an integrated plastic base plate.

II. SYSTEM ARCHITECTURE

The OTS sensor is a reflective-type intensity modulatedfiber-optic sensor (FOS) with a built-in light emitting diode(LED), and a flexible 10 m plastic-sheathed fiber-optic cable(FUE 500C1003, Baumer Electric, Switzerland). The FOShas a visible red light (centered at 680 nm) emitter elementand a receiving element that are mounted side by side in a

cylindrical brass housing (outer diameter 4.5 mm, length 24mm) placed behind a custom-built plano-concave lens (focallength -10 mm and the radius of the lens aperture 5.0 mm).The lens is made of poly(methyl methacrylate) (PMMA,Plexiglas, acrylic glass). When a beam of light is generatedby the LED, it propagates down the optical fiber and emitsinto the air at the fiber tip. It then passes through the lens intoan acrylic prism that is cast into a wedge-shaped structurewith a dimension of 70mm 10mmT mm (with T from20 to 2). The wedge prism is designed to have non-parallelfront and back faces that are separated by a small angle a =15o. The plano-concave lens is a divergent optical elementand provides the required light beam profile by expandingthe area covered by the LED. FOS receiving element receivesreflections of light that become trapped within the prism.A rectangular plastic base plate (90mm 70mm 15mm)with a narrow slot in the middle (85mm 10mm 15mm)made of polyacetal copolymer (POM-C, Angst + Pfister,Switzerland) along with a polyamide M6 hex nut (Richco,USA) secures the prism, lens, and FOS in place as shownin Fig. 1 (bottom). The light beams illuminating the insideof the wedge prism through the lens can be seen in Fig.1 (top panel). The technical specifications of the FOS aresummarized in Table 1.

TABLE 1SPECIFICATIONS OF THE OTS SENSOR

Parameter type Specification

Sensing distance 25 mmLight source Pulsed red LEDWave length 680 nmVoltage supply range 10.8 to 26.4 VDCAnalog voltage output 1.0 to 5.0 VDC

FOS is connected to a circuit board (Fig. 2) placed outsidethe MRI room via a 10 m fiber-optic cable. This unitconsists of a 15V AC-to-DC converter (PM10-15, MeanWell Inc., USA), a photodiode with an analog voltage output(FWDK 10U84Y0, Baumer Electric, Switzerland), and a dataacquisition (DAQ) card (USB-6008, National Instruments,USA). The DAQ card is used as interface between theoptical sensing unit and a computer on which the operationalsoftware runs in conjunction with the OTS.

Figure 2: 3D simulation of the handheld response system showing the behaviour of light beams in prism. The total reflecting surface is used as slider stripe (black arrow). On the opposite side strong refraction of light take place (dashed arrow) .

Frustrated internal total reflection takes place by changing the refractive indices on the slider stripe

during finger movements. This fact leads to a decreasing number of internal total reflections and

an increasing received light intensity, which is collected on the light receiver and processed via the

analog amplifier in Fig. 3.

System Architecture

Figure 3: Handheld response circuit board including power supply, data acquisition system and analog amplifier. Note that USB-6008 uses the 12 bit differential analog input mode AI+ and AI-, because the signal to the device is low level and requires greater accuracy.

+Vs analog FWDK 0 V 10U84Y0

Analog Amplifier + 15V

PM10-15 0 V

AC/DC Converter

230 VAC

AI+

NI USB-6008 AI-

0 V

10 k

100 nF

USB

DAQ USB Device

FOS

Fig. 2. OTS circuit board including a power supply, data acquisition system,and optical amplifier. A low-pass filter (R= 10 k, C= 100 nF) with a cut-offfrequency fc = 159 Hz was used to minimize noise.

III. PRINCIPLE OF OPERATION

The principle of operation of the prototype OTS is basedon the frustrated total internal reflection (FTIR) caused bythe difference in the refractive indices of the adjacent media.According to Snells Law [17], when light passes from onemedium to another with different refractive indices such thatn1 > n2, total reflection of the incoming beam can occur atthe interface between the two media if the angle of incidentlight beam (qi) exceeds a critical angle (qc) defined by (1):

qi > qc := arcsin(n2/n1) (1)

Fig. 3. 2D simulation of light beam propagation in an acrylic wedge prismthrough a plano-concave lens. The top surface acts as a touch-sensitive areawhere FTIR takes place while strong refraction occurs on the opposite face.

In the case of a light beam traveling through an acrylicwedge prism (n1 = 1.5) into air (n2 = 1), the critical anglefor the total internal reflection to occur at the acrylic-airboundary is qc = 41.8o. Prisms can deviate light beamsdepending on how they are oriented relative to the inputlight, and how the prisms are cut. When the incident lightbeam enters an acrylic prism at an angle greater than itscritical angle (i.e. qi > 41.8o), no refraction occurs in thewedge prism, and the light is totally internally reflected.By fabricating non-parallel front and back faces for thewedge prism that are separated by a small angle, totalinternal reflection of the light inside the wedge prism canbe interrupted. An optimal range of a = 8.8o to 23o wasestimated for wedge angle by running simulations withthe WinLens 3D Basic software suite (Qioptiq, Germany).Sample 2D and 3D simulations are shown in Figs. 3 and 4. Afinal wedge angle a = 15o was determined experimentally tobe the best based on the characteristics of the FOS (Table 1),lens numerical aperture [18], and manufacturing techniques.

Fig. 4. 3D simulation of light beam propagation in an acrylic wedge prismthrough a plano-concave lens. FTIR phenomenon occurs at the top surfaceof the wedge prism (i.e. the touch stripe, indicated by black arrow) whilestrong refraction occurs in the opposite face (indicated by dashed arrow).

This method traps the light inside the acrylic wedge prism,which is frustrated (scattered) at the point of a touch [11],[19]. The scattered light is then collected by a photodetector.As shown in Fig. 3, qik qi1 a(k 1) qc with thenumber of total reflections k= {1,2,3, ...}, a 2 [8.8o , 23o],the reflection angle (qik ) decreases with increasing k. Thus,touching the surface of the acrylic wedge prism where FTIRphenomenon takes place (i.e. on the touch stripe), at adistance further away from the FOS leads to an increasein total internal reflections. This, in turn, generates greateroutput voltage signals that can be used to differentiate theposition of fingertip on the touch sensitive stripe area.

IV. OPERATIONAL SOFTWARE

The real-time operational software is implemented byusing the National Instruments LabViewTM software package(TX, USA). Custom-written software is run on the PCnext to the MRI scanner to acquire output signals, performsignal analysis, and display user interface data on graphicsmonitor appropriate to a particular user task. For example,the software can translate the differences in output voltagefrom touching the OTS at different distances (Fig. 5 (a), (b))as the position of a sliding piece in a visual analog scale(Fig. 5 (c), (d)) or the pressing of the buttons (Fig. 5 (e),(f)). Fig. 6 shows LabVIEW block diagram that generatesthe visual-analog scale, which is widely used in fMRI as agraphic rating scale, as an example of an interactive display.

Fig. 5. (a) Difference in fingertip touch position results in different outputvoltage which is translated by the software to (c),(d) a position of a slidingpiece on the visual analog rating scale; or (e),(f) a pressed push button.

Fig. 6. Example of LabVIEW block diagram for visual-analog rating scale

V. PERFORMANCE EVALUATIONTo evaluate the accuracy of the fingertip position on

the touch-sensitive stripe, we measured the intensity of theresultant internally-reflected light beams that were capturedby the FOS receiving element and converted into an electricalsignal by the photodiode. The amplified and low-pass filtered(refer to Fig. 2) output signal from the photodiode wasmeasured with a digital multimeter (HP34401A, Hewlett-Packard, CO, USA). Fig. 7 shows a graph of percent outputvoltage versus position (distance) of the fingertip with respectto the FOS receiving element. The nonlinearity of the curve iscaused by manufacturing tolerances of the handmade prism.The spatial resolution is obtained in the middle distancerange by determining the smallest distinguishable outputvoltage difference (in this case 10 mV) at two adjacent touchpositions. The value obtained is about 0.5 cm, which iscomparable with the resolutions of conventional touchpadsystems, ensuring even close-distance detection of touchsignals from the users fingertip.

0

20

40

60

80

100

0 1 2 3 4 5 6 7 8

% O

utp

ut

Vo

lta

ge

[V

]

Touch Distance [cm]

% Output Voltage vs. Touch Distance

Measurement 1 Measurement 2 Measurement 3

Fig. 7. Evaluation of the accuracy of fingertip touch position on the stripe

VI. MRI COMPATIBILITY TESTSA cylindrical homogeneous water phantom filled with 1.7

gram/liter of nickel chloride and 7 gram/liter of sodiumchloride underwent echo-planar imaging (EPI) to investigatewhether introducing the OTS to the MRI setting interferedwith the signal during scanning. The phantom was scannedunder four testing conditions: (a) without the presence of theOTS (i.e. baseline), (b) with the OTS present but turned off,(c) with the OTS present and turned on but not actuated,and (d) with an actuated OTS. All images were acquired ona clinical 3.0 T whole-body MRI system (Ingenia, PhilipsHealthcare, Best, The Netherlands) with 8-channel head-coilat the MRI center of the University Hospital Zurich. Imageswere acquired in an interleaved fashion with thirty four axialslices (each 3 mm thick, 1 mm interslice gap) using a T2-sensitive EPI scan with a flip angle = 80, repetition time(TR) = 2000 ms, echo time (TE) = 30 ms, field of view(FOV) = 240 240 mm2, image matrix = 96 96, voxelsize = 3 3 3 mm3. We calculated the time-variant signal-to-noise ratio (tSNR) as a measure of image time coursestability and degree of signal distortion (Fig. 8). tSNR isdefined as the ratio between the temporal mean of a timeseries xi and its temporal standard deviation [20]:

tSNR=s

=q

1N

Ni=1 (xi)2

(2)

where N is the number of time points, is the mean ofthe time series and s is its standard deviation. As shownin Fig. 9, overall tSNR was decreased by no more than 3%across different fMRI compatibility test conditions.

Not present

Off On Actuated

Mean 70.37 69.88 69.50 68.33 SD 1.750 1.642 1.604 1.527

66

67

68

69

70

71

72

73

tSN

R

Fig. 8. Average tSNR for different fMRI compatibility test conditions

0.00

0.69

1.24

2.90

0

1

2

3

4

Not present Off On Actuated

% t

SN

R L

oss

Fig. 9. Percent tSNR loss for different fMRI compatibility test conditions

We also evaluated the MR image quality with imagesubtraction method for each condition (Fig. 10). Changesfrom baseline (the phantom without the presence of OTS)were assessed by subtracting the absolute value betweenphantom image for that condition from the baseline image.Furthermore, upon visual inspection of all images andfield maps, no noticeable image artifact associated withthe presence of the device was observed. Likewise, nomagnetically induced force or torque was observed whenOTS was placed inside the scanner bore (about 50 cm awayfrom the phantom base) before and during the fMRI scans.

Fig. 10. (Top) A graphical representation of the imaging results;(Bottom) Image subtraction results with condition (a) as a baseline.

VII. CONCLUSIONWe presented the design and architecture of a novel fMRI-

compatible optical touch stripe as a feasible user interfaceto emulate physical push buttons or sliding bars duringfMRI experiments. By taking advantage of a frustrated totalinternal reflection concept and using an intensity-modulatedfiber optic sensor, our device can accurately detect a touchsignal from a user inside a 3.0 Tesla MRI scanner without aneed for an IR sensor, an imaging camera, or any electroniccomponents in close vicinity of the magnet.

REFERENCES[1] B. Jarrahi, J. Wanek, U. Mehnert, and S. Kollias, An fmri-compatible

multi-configurable handheld response system using an intensity-modulated fiber-optic sensor, in Engineering in Medicine and BiologySociety (EMBC), 2013 35th Annual International Conference of theIEEE. IEEE, 2013, pp. 63496352.

[2] Apple iPhone, Available at https://www.apple.com/iphone/, [Lastaccessed on March 15th, 2008].

[3] Apple iPod touch, Available at https://www.apple.com/ipod/, [Lastaccessed on March 15th, 2008].

[4] Microsoft Surface Tablets - The Windows Tablet That Does More,Available at http://www.microsoft.com/surface, [Last accessed onMarch 15th, 2008].

[5] SMART Table442i collaborative learning center, Available athttp://smarttech.com/table, [Last accessed on March 15th, 2008].

[6] R. Bourdelais, R. Cok, and C. Kaminsky, Flexible resistive touchscreen, Oct. 28 2004, united States Patent App. 10/422,583. [Online].Available: http://www.google.com/patents/US20040212599

[7] R. Adler and P. J. Desmares, An economical touch panel using sawabsorption, Ultrasonics, Ferroelectrics and Frequency Control, IEEETransactions on, vol. 34, no. 2, pp. 195201, 1987.

[8] R. S. Dahiya, G. Metta, M. Valle, A. Adami, and L. Lorenzelli,Piezoelectric oxide semiconductor field effect transistor touch sensingdevices, Applied Physics Letters, vol. 95, no. 3, p. 034105, 2009.

[9] P. Kalendra and W. Piazza, Automatic calibration of a capacitivetouch screen used with a fixed element flat screen display panel,Feb. 1 1994, united States Patent 5,283,559. [Online]. Available:https://www.google.com/patents/US5283559

[10] R. Doering, Infrared touch panel, Sept. 19 1989, unitedStates Patent 4,868,912. [Online]. Available: https://www.google.com/patents/US4868912

[11] J. Y. Han, Low-cost multi-touch sensing through frustrated totalinternal reflection, in Proceedings of the 18th annual ACM symposiumon User interface software and technology. ACM, 2005, pp. 115118.

[12] G. Schaefers, Testing mr safety and compatibility: an overview ofthe methods and current standards. IEEE engineering in medicineand biology magazine, vol. 27, no. 3, p. 23, 2008.

[13] G. Schaefers and A. Melzer, Devices and materials in mri, SpringerHandbook of Medical Technology, pp. 503521, 2012.

[14] R. Frayne, B. Goodyear, P. Dickhoff, M. Lauzon, and R. Sevick,Magnetic resonance imaging at 3.0 tesla: challenges and advantagesin clinical neurological imaging, Investigative radiology, vol. 38,no. 7, pp. 385402, 2003.

[15] S. Hodges, S. Izadi, A. Butler, A. Rrustemi, and B. Buxton, Thinsight:versatile multi-touch sensing for thin form-factor displays, inProceedings of the 20th annual ACM symposium on User interfacesoftware and technology. ACM, 2007, pp. 259268.

[16] F. Tam, N. W. Churchill, S. C. Strother, and S. J. Graham, A newtablet for writing and drawing during functional mri, Human brainmapping, vol. 32, no. 2, pp. 240248, 2011.

[17] M. Born and E. Wolf, Principles of optics: electromagnetic theory ofpropagation, interference & diffraction of light. CUP Archive, 1999.

[18] H. P. Zappe, Fundamentals of Micro-optics. Cambridge UniversityPress Cambridge, 2010, vol. 90.

[19] J. Schoning, P. Brandl, F. Daiber, F. Echtler, O. Hilliges, J. Hook,M. Lochtefeld, N. Motamedi, L. Muller, P. Olivier, et al., Multi-touchsurfaces: A technical guide.

[20] R. Weisskoff, Simple measurement of scanner stability for functionalnmr imaging of activation in the brain, Magnetic resonance inmedicine, vol. 36, no. 4, pp. 643645, 2005.